Skip to main content

Full text of "DTIC ADA579182: Development of a Design Tool for Planning Aqueous Amendment Injection Systems Permanganate Design Tool"

See other formats


FINAL REPORT & USER’S GUIDE 


Development of a Design Tool for Planning Aqueous 
Amendment Injection Systems 
Permanganate Design Tool 

ESTCP Project ER-200626 


August 2010 


Robert Borden 
Ki Young Cha 

North Carolina State University 

Thomas Simpkin 

CH2M HILL, Inc. 

M. Tony Lieberman 

Solutions-IES, Inc. 


This document has been cleared for public release 





Report Documentation Page 

Form Approved 

OMB No. 0704-0188 

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and 
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, 
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington 

VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it 
does not display a currently valid OMB control number. 

1. REPORT DATE 

AUG 2010 2. REPORT TYPE 

3. DATES COVERED 

00-00-2010 to 00-00-2010 

4. TITLE AND SUBTITLE 

Development of a Design Tool for Planning Aqueous Amendment 

Injection Systems Permanganate Design Tool 

5a. CONTRACT NUMBER 

5b. GRANT NUMBER 

5c. PROGRAM ELEMENT NUMBER 

6. AUTHOR(S) 

5d. PROJECT NUMBER 

5e. TASK NUMBER 

5f. WORK UNIT NUMBER 

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 

North Carolina State University,2101 Hillsborough 

Street,Raleigh,NC,27695 

8. PERFORMING ORGANIZATION 

REPORT NUMBER 

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 

10. SPONSOR/MONITOR’S ACRONYM(S) 

11. SPONSOR/MONITOR’S REPORT 
NUMBER(S) 

12. DISTRIBUTION/AVAILABILITY STATEMENT 

Approved for public release; distribution unlimited 

13. SUPPLEMENTARY NOTES 

14. ABSTRACT 

15. SUBJECT TERMS 

16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 

ARSTRATT 

18. NUMBER 19a. NAME OF 

OF PAGES RESPONSIBLE PERSON 

a. REPORT b. ABSTRACT c. THIS PAGE Same aS 

unclassified unclassified unclassified Report (SAR) 

90 


Standard Form 298 (Rev. 8-98) 

Prescribed by ANSI Std Z39-18 





This report was prepared for the Environmental Security Technology Certification Program 
(ESTCP) by North Carolina State University (NCSU) and representatives from ESTCP. In no 
event shall either the United States Government or NCSU have any responsibility or liability for 
any consequences of any use, misuse, inability to use, or reliance upon the information contained 
herein, nor does either warrant or otherwise represent in any way the accuracy, adequacy, 
efficacy, or applicability of the contents hereof To discuss applications of this technology please 
contact: 

Dr. Robert C. Borden of North Carolina State University can be reached by phone at 919-515- 
1625 or by email at rcborden@eos.ncsu.edu 



ACKNOWLEDGEMENTS 


We gratefully acknowledge the financial and technical support provided by the Environmental 
Technology Certification Program including the guidance provided by Dr. Andrea Leeson, Erica 
Becvar (the Contracting Officer’s Representative), and Dr. Marvin Unger (ESTCP reviewer). We 
would also like to thank the members of the Technical Advisory Committee and the ER-0623 
project team whose work greatly improved the quality and usefulness of the design tool. 



ACRNOYM LIST 


CC1 4 

CS-10 

CSTR 

CT 

EDB 

ESTCP 

ft Bgs 
gP m 

ISCO 

KMn0 4 

MCL 

ME 

MinOx 

MMR 

M 11 O 4 

NaMn0 4 

NOD 

O&M 

ODE 

OF 

PCE 

RMSE 

ROI 

SF V 

SSES 

TCE 

UOD 

uscu 


carbon tetrachloride 

Chemical Spill 10 
continuously stirred tank reactors 
contact time 

ethylene dibromide 

Environmental Security Technology Certification Program 

feet below ground surface 

gallons per minute 

in situ chemical oxidation 

potassium permanganate 

maximum contaminant levels 
mean error 

minimum oxidant concentration 
Massachusetts Military Reservation 
permanganate 

sodium permanganate 
natural oxidant demand 
operation and maintenance 

ordinary differential equations 
overlap factor 
perchloroethene 

root mean square error 
radius of influence 

volume scaling factor 
simple scoring error statistic 

trichloroethene 

Ultimate Oxidant Demand 
unified soil classification system 


iii 



EXECUTIVE SUMMARY 


In Situ Chemical Oxidation (ISCO) with permanganate (MnOzj) has been applied at hundreds of 
sites to treat aquifers contaminated with chlorinated ethenes and other contaminants. In this 
process, a Mn04 solution is injected into the subsurface using temporary points or permanent 
wells. Once injected, the Mn04 is transported through the aquifer by ambient or induced 
groundwater flow. Major capital costs associated with the process include: (a) purchase of the 
chemical reagent (e.g. permanganate); (b) installation of injection points or wells; and (c) labor 
and equipment to implement the injection. 

For ISCO to be effective, the permanganate must contact the contaminant. This can be difficult 
in many aquifers because natural heterogeneities can result in flow bypassing around lower 
permeability zones. A variety of methods can be used to provide more effective reagent 
distribution including injecting more reagent, injecting more water to distribute the reagent and 
using more closely spaced injection points. However, all of these approaches increase costs. In 
this project, we have developed a design tool to assist users in developing more effective and less 
costly permanganate injection systems. 

An ISCO reaction module for the RT3D numerical model was first developed. The reaction 
between Mn04 and a single contaminant is simulated as an instantaneous reaction. Mn04 
consumption by the natural oxidant demand (NOD) is modeled assuming NOD is composed of 
two fractions: NODi which reacts instantaneously with permanganate; and NODs which reacts 
with permanganate by a 2nd order relationship. The newly developed model was then evaluated 
by comparing model simulation results with field monitoring date from an ISCO pilot test 
conducted at the Massachusetts Military Reservation. Kinetic parameters used to calibrate the 
model were estimated from prior laboratory tests. 

The ISCO model was then applied to a hypothetical heterogeneous aquifer to evaluate the effect 
of different design variables and aquifer parameters on treatment efficiency. Model simulation 
results indicate that ISCO performance is most sensitive to: (1) mass of permanganate injected; 
and (2) volume of water injected. Reducing the injection wells spacing and performing multiple 
injections had less benefit when the volume and concentration of Mn04 solution was held 
constant. Model sensitivity analyses indicated that ISCO performance was sensitive to the 
kinetics of Mn04 consumption by NOD and it is probably not feasible to develop a simple set of 
design curves relating distribution efficiency to amount of reagent and water injected. 

An Excel spreadsheet based design tool (CDISCO) was developed to assist users in the design of 
ISCO systems with Mn04. Comparisons with analytical and numerical models demonstrated 
that CDISCO provides reasonably good estimates of the average Mn04 transport distance in 
heterogeneous aquifers. However, CDISCO will under estimate the maximum Mn04 transport 
distance in higher permeability layers. The primary model inputs for CDISCO are the aquifer 
characteristics, injection conditions, unit costs, and a Radius of Influence (ROI) overlap factor 
(OF). Comparison with the 3D simulations also showed that values of OF between 1.0 and 1.5 
will generally result in good remediation system performance. 


IV 



Table of Contents 


1.0 INTRODUCTION-1 

1.1 Background-1 

1.2 Project Objectives-1 

1.3 Stakeholder / End-User Issues-2 

2.0 TECHNOLOGY DESCRIPTION - ISCO WITH PERMANGANATE -3 

2.1 Introduction-3 

2.2 Procedures for Injecting Permanganate-4 

2.2.1 Arrangement of Injection Points-4 

2.2.2 Injection Point Construction-5 

2.2.3 Amount of Water and Chemical Reagent to Inject-5 

2.2.4 Reinjection Frequency-5 

2.2.5 Additional Labor and Equipment Required-6 

3.0 PERMANGANATE CONSUMPTION BY AQUIFER MATERIAL-7 

3.1 Introduction-7 

3.2 Fate and Transport of Permanganate in the Subsurface-7 

3.2.1 Natural Oxidant Demand (NOD)-7 

3.2.2 Modeling Approaches-9 

3.3 Kinetic Model Evaluation-9 

3.3.1 Model 1 - Zero Order Loss of Mn 04 -11 

3.3.2 Model 2 - First Order Loss of Mn 04 -12 

3.3.3 Model 3 - First Order Loss of NOD-13 

3.3.4 Model 4 - Second Order Loss of Mn 04 and NOD-15 

3.3.5 Model 5 - Second Order Loss of Mn 04 with Fast and Slow NOD-16 

3.3.6 Model 6 - Second Order Loss of Mn 04 with Instantaneous and Slow NOD — 18 

3.3.7 K i netic Model Evaluation Summary-19 

3.4 MMR Parameter Estimates-20 

3.5 Summary and Conclusions - Mn 04 Consumption by Aquifer Material-22 

4.0 MODEL TESTING - MMR ISCO PILOT TEST-24 

4.1 Introduction-24 

4.2 Massachusetts Military Reservation (MMR)-24 

4.3 Pilot Test-26 

4.4 Modeling of MMR Pilot Test-32 

4.4.1 Reaction K in etics-32 

4.4.2 Numerical Implementation-33 

4.4.3 Model Setup-33 

4.5 Model Calibration-39 

4.5.1 Simple Scoring Error Statistics (SSES)-39 

4.5.2 Model Calibration Results-40 

4.6 Summary and Conclusions - MMR Model Evaluation-42 

5.0 EFFECT OF INJECTION SYSTEM DESIGN ON PERFORMANCE - 43 

5.1 Introduction-43 

5.2 Model Setup and Base Case Conditions-43 

5.2.1 Scaling Factors-46 

5.2.2 Typical Simulation Results-47 


v 












































5.2.3 Treatment Efficiency Criteria-50 

5.3 Effect of Fluid Volume, Permanganate Mass and Time on 

Treatment Efficiency-50 

5.4 Effect of Injection Design Parameters on Performance-53 

5.5 Effect of Site Characteristics on Performance-55 

5.5.1 Effect of Aquifer Heterogeneity on Em-59 

5.6 Summary and Conclusions - Effect of Injection System Design Variables 

and Site Characteristics on Remediation System Performance-61 

6.0 SPREADSHEET BASED MODELING OF PERMANGANATE 

DISTRIBUTION-63 

6.1 Introduction-63 

6.2 Simulating Oxidant Distribution Using a Series of CSTRs-63 

6.2.1 Model Development-63 

6.2.2 Model Validation-65 

6.3 Comparison of CDISCO with 3D Heterogeneous Simulations-68 

6.4 CDISCO Model Structure-71 

6.5 Effect of Overlap Factor on Contact Efficiency-73 

6.6 Summary-74 

7.0 REFERENCES-75 

8.0 POINTS OF CONTACT-80 


vi 




















LIST OF TABLES 


Table 3.1: 
Table 3.2: 
Table 3.3: 
Table 3.4: 
Table 3.5: 
Table 3.6: 
Table 3.7: 
Table 3.8: 
Table 3.9: 
Table 3.10: 
Table 3.11a: 
Table 3.11b: 

Table 3.11c: 
Table 3.12: 
Table 4.1: 
Table 4.2: 
Table 4.3: 
Table 4.4: 
Table 4.5: 
Table 4.6: 
Table 4.7: 
Table 4.8: 

Table 5.1: 
Table 5.2: 
Table 5.3: 

Table 5.4: 
Table 6.1: 

Table 6.2: 


Reported Values of NOD for Different Sites 
Batch Experimental Conditions of Each Treatment 
Statistical Results of Model 1 Evaluation 
Statistical Results of Model 2 Evaluation 
Statistical Results of Model 3 Evaluation 
Statistical Results of Model 4 Evaluation 
Statistical Results of Model 5 Evaluation 
Statistical Results of Model 6 Evaluation 
Best Fit Coefficients for Model 4, 5, and 6 
MMR Soil Sample Comparison 

Best Fit Parameter Estimates for MMR Soils - Total NOD (mmol/g) 

Best Fit Parameter Estimates for MMR Soils - Slow Reaction Rate (K. 2 s) 
(L/mmol-d) 

Best Fit Parameter Estimates for MMR Soils - Fracation Instantaneous 

Parameter Set for MMR 

Well Construction Information 

TCE Monitoring Results 

Permanganate Monitoring Results 

Injection Flow Rates and Concentrations Used in Model Simulations 
List of Common Parameters Used in Calibration Model 
Details of 4 Simulation Scenarios 

Simulated and Observed Contaminant (TCE) and Mn 04 Error Statistics 
Error Statistics Comparing Simulated and Observed Permanganate Measurements 
with Increased Total NOD 
Base Case Simulation Conditions 

Target Characteristics for Low, Moderate and High Levels of Heterogeneity. 
Statistical Characteristics of Ln Transformed Hydraulic Conductivity 
Distributions used in Model Simulations 
Input Parameters used in Sensitivity Analyses Simulations 
Base Model Parameters for Comparison of CDISCO, Analytical and 
RT3D Simulations 

Comparison of CDISCO, Analytical and RT3D Non-Reactive Simulations 


vii 






LIST OF FIGURES 


Figure 3.1: 

Figure 3.2: 

Figure 3.3: 

Figure 3.4: 

Figure 3.5: 

Figure 3.6: 

Figure 4.1: 
Figure 4.2: 

Figure 4.3: 

Figure 4.4: 
Figure 4.5: 
Figure 4.6: 
Figure 4.7: 
Figure 4.8: 
Figure 4.9: 
Figure 4.10 

Figure 5.1: 
Figure 5.2: 
Figure 5.3: 


Figure 5.4: 


Figure 5.5: 


Figure 5.6: 
Figure 5.7: 
Figure 5.8: 
Figure 5.9: 


Comparison of Observed Values of AMn 04 with Model 1 Simulation Results (all 
data for Soil C) 

Comparison of Observed Values of AMn 04 with Model 2 Simulation Results (all 
data for Soil C) 

Comparison of Observed Values of AMn 04 with Model 3 Simulation Results (all 
data for Soil C) 

Comparison of Observed Values of AMn 04 with Model 4 Simulation Results (all 
data for Soil C) 

Comparison of Observed Values of AMn 04 with Model 5 Simulation Results (all 
data for Soil C) 

Comparison of Observed Values of AMn 04 with Model 6 Simulation Results (all 
data for Soil C) 

Location of MMR on Cape Cod, Massachusetts 

Plume Distribution of MMR (grey area represent MMR, red line represent plume 
boundary, AFCEE 2007b) 

CS-10 Plume (grey area represent MMR, red line represent plume boundary, 
AFCEE 2007b) 

Cross Section of Pilot Test Area (CH2M Hill 2007) 

Plan-View of MMR Pilot Test Simulation Grid 
Profile-View of MMR Pilot Test Simulation Grid 
Cross-Section View of Permeability Distribution 

Plan-View of 15th Layer of Model Showing and Injection and Monitoring Wells 
Front View of 50th Row of Model Showing Injection and Monitoring Wells 
Pro file-View of Contaminant and Permanganate Distribution at 6, 18, 30 and 90 
Days of Simulation with Scenario 4 (deep red indicate high concentration) 
Hypothetical Injection Grid Showing Model Domain Subarea 
Model Domain for Base Case Condition. 

Horizontal Hydraulic Conductivity, MnCL, NODi, NODs and Contaminant 
Distributions in Top Layer of Aquifer at 120 days after Injection for Moderately 
Heterogeneous Aquifer when Wells 1-5 are Injected with SF V = SF M = 0.25 
Horizontal Hydraulic Conductivity, MnCL, NODi, NODs and Contaminant 
Distributions in Last Row of Aquifer at 120 days after Injection for Moderately 
Heterogeneous Aquifer when Wells 1-5 are Injected with SFy = SF M = 0.25 
Variation in Aquifer Volume Contact Efficiency and Contaminant Mass 
Treatment Efficiency (Em) with Time where Fluid Injection Volume is held 
Constant (SF V =0.25) and Mn 04 Mass Varies (SF M varies from 0.1 to 1.0) 
Variation in Aquifer Volume Contact Efficiency (Ey) and Fraction Unreacted 
MnCL (U) at 180 days after Injection with Mass and Volume Scaling Factors 
Effect of Well Spacing on E v and U at 180 days for SF V = SF M in a Medium 
Heterogeneity Aquifer 

Effect of Reinjection on E v and U at 180 days for SFy = SFm in a Medium 
Heterogeneity Aquifer. Well spacing = 3 m. 

Effect of NOD Kinetic Parameters on E v and U at 180 days for SF V =SF M : (a) 
Slow NOD Reaction Rate (K 2 s); (b) Total NOD; and (c) Fraction NODi. 


viii 



Figure 5.10 
Figure 5.11 
Figure 6.1: 
Figure 6.2: 
Figure 6.3: 
Figure 6.4: 
Figure 6.5: 

Figure 6.6: 
Figure 6.7: 
Figure 6.8: 


Effect of Initial Contaminant Concentration (a) and Contaminant Retardation 
factor (b) on Ey, Em and U at 180 days for SFv=SFm- 

Effect of Low, Medium and High Aquifer Heterogeneity on E v , E M and U at 180 
days (SFy = SFm) 

Comparison of CDISCO, RT3D and ID Analytical Solutions of Non-Reactive 
Solute Transport Away from a Single Injection Well at 1 day after Injection 
Comparison of CDISCO and RT3D Simulations of Mn 04 , NODi, NODs and 
Contaminant Concentration at 10 days after Injection for oil =0.1 m. 
Non-Reactive Solute Concentrations versus Radial Distance from Injection Well 
Generated in 3D Heterogeneous RT3D Simulation. 

Comparison of CDISCO Simulation (oil = 1.5 m) and Spatially Averaged 
Concentrations from 3D RT3D Simulation 

Comparison of Model Results at 10 days after Injection for 1-D Homogeneous 

CDISCO Simulation and 3-D spatially Heterogeneous RT3D Simulation for: (a) 

MnO/t; (b) NODi; (c) NODs and Contaminant 

Typical Output from Permanganate Transport Model 

Typical Output from Injection Scenario Cost Comparison 

Effect of Overlap Factor (OF) on Aquifer Volume Contact Efficiency (Ey) and 

Contaminant Mass Treatment Efficiency (E M ). 


IX 



1.0 


INTRODUCTION 


1.1 BACKGROUND 

In situ chemical oxidation (ISCO) using permanganate (MnOzj) can be effective for in situ 
treatment of chlorinated ethenes and other groundwater contaminants if the M11O4 contacts the 
target contaminant. 

There are a variety of different approaches that can be used to distribute Mn 04 in the subsurface 
including: (a) injection only using grids of temporary or permanent wells; and (b) recirculation 
using systems of injection and pumping wells. Each of these approaches has advantages and 
disadvantages with the ‘best’ approach dependent on site-specific conditions. For each 
approach, cost and effectiveness are a function of the well layout and injection sequence. 
Consequently, the ‘optimum design’ will include a specific arrangement of wells, injection 
volumes and rates, and amount of reagent. Existing guidance documents (ITRC 2005, and 
Huling and Pivetz 2006) provide general information on how the remediation process works and 
factors to consider when planning an injection system. However, these documents do not 
provide specific information on how to actually design an injection system to provide good 
amendment distribution at a reasonable cost. 

In recent years, a number of computer modeling packages have been developed that can be used 
to simulate the reactive transport under reasonably realistic (i.e. heterogeneous) conditions. With 
these tools, users can evaluate alternative injection approaches and identify the ‘best’ design 
based on site-specific conditions including aquifer permeability and heterogeneity, contaminant 
distribution, site access limitations, drilling, labor and material costs, etc. 

Unfortunately, these models are only rarely used. In most cases, remediation systems are 
designed by based on rules of thumb and prior experience. Sometimes this approach results in a 
good, efficient design. However in some cases, designs are less effective than desired and more 
expensive than needed. To reduce remediation system costs and improve performance, tools are 
needed that allow engineers to quickly identify an efficient design for the specific conditions at 
their site without extensive site characterization and a high level of modeling expertise. 

1.2 PROJECT OBJECTIVES 

The overall objective of this project is to develop a tool to assist in the design of in situ chemical 
oxidation systems using permanganate. Specific objectives of this project are listed below. 

1. Using currently available numerical models, examine the effects of site conditions (e.g. 
permeability, contaminant distribution, site heterogeneity) and design variables (location 
of wells, injection rates, volumes, amount of reagent, etc.) on permanganate distribution 
and associated contact efficiency. If possible, develop simple design curves relating 
distribution efficiency to amount of reagent and water injected. Determine if there are 
significant differences in performance between different injection patterns. If possible, 
present the results in a normalized or non-dimensional form. The information learned 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


1 


August 2010 



from the modeling will provide guidance to design tool users in selection of important 
design parameters (e.g., pore volumes of injection fluid, amount of reagent, etc.). 

2. Develop a simple, spreadsheet-based tool to assist in the design of MnC^ injection 

systems. This design tool will allow designers to evaluate the effect of different variables 
(well spacing, amount of reagent and water, injection rate, etc.) on remediation system 
cost and expected performance. Experienced users who have already compiled the input 
data for their site (e.g. permeability, NOD, contaminant concentrations) should be able to 
quickly develop and evaluate several alternative designs. 

1.3 STAKEHOLDER / END-USER ISSUES 

The primary objective of this project was to develop a design tool that is easy to learn, simple to 
use, and widely applied. The design tool is structured to allow new users to download the 
required materials, and complete a preliminary injection system design in a few hours without 
extensive groundwater modeling experience. However, users are expected to be familiar with 
basic fundamentals of groundwater flow, solute transport, and ISCO using Mn 04 . The design 
tool and guidance document are available for download from one or more websites. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


2 


August 2010 



2.0 


TECHNOLOGY DESCRIPTION - ISCO WITH PERMANGANATE 


2.1 INTRODUCTION 

Potassium permanganate (KMnO/t) and sodium permanganate (NaMnO/t) have been injected at 
hundreds of sites to chemically oxidize chlorinated solvents and other ground water 
contaminants. In this process, a permanganate (MnO/t) solution is injected into the subsurface 
using temporary points or permanent wells. Once injected, the Mn 04 is transported through the 
aquifer by ambient or induced groundwater flow. Major capital costs associated with the process 
include: (a) purchase of the chemical reagent (e.g. permanganate); (b) installation of injection 
points or wells; and (c) labor and equipment to implement the injection. 

Permanganate is most commonly injected into the subsurface through a grid of wells. To be 
effective, the chemical reagent must be brought into close contact with the contaminants to be 
treated. This can be difficult in many aquifers because natural heterogeneities can result in flow 
bypassing around lower permeability zones. A variety of methods can be used to provide more 
effective reagent distribution including injecting more reagent, injecting more water to distribute 
the reagent and using more, closely spaced injection points. However, all of these approaches 
increase costs. 

In this project, a design tool is developed to assist users in designing MnC >4 injection and 
distribution systems. Prior to using the design tool, users should: (1) have a good understanding 
of the ISCO process; and (2) have completed a preliminary screening to determine if ISCO with 
Mn 04 is appropriate for their site. For information on ISCO with Mn 04 , consult the following 
documents. 

• In-Situ Chemical Oxidation - Engineering Issue, by S. G. Huling and B. Pivetz. US 
Environmental Protection Agency, National Risk Management Research Laboratory, R.S. 
Kerr Environmental Research Center, Ada, Oklahoma. EPA/600/R-06/072, 2006. 
(http://www.epa.gov/ada/download/issue/600R06072.pdf) . 

• Technical and Regulatory Guidance for In Situ Chemical Oxidation of 
Contaminated Soil and Groundwater 2nd Ed., by the Interstate Technology & 
Regulatory Council, Washington, D.C., 2005. (http://www.itrcweb.org/gd ISCO. asp) . 

• Decision Support Tools for In Situ Chemical Oxidation, by R. L. Siegrist, M.L. Crimi, 
B. Petri, T. Simpkin, T. Palaia, F.J. Krembs, J. Munakata-Marr, T. Illangasekare, G. Ng, 
M. Singletary, and N. Ruiz. 2009. Final project report to the U.S. Environmental 
Security Technology Certification Program for ESTCP project ER-0623.ools. 
(http://docs.serdp-estcp.org) 

Permanganate has been applied at hundreds of commercial and military sites. Although this 
process has been demonstrated in the laboratory and the field, the technology continues to 
evolve. This design tool is based on the current state of practice at the time of writing. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


3 


August 2010 





2.2 PROCEDURES FOR INJECTING PERMANGANATE 


ISCO projects using permanganate typically, but not always, involve the following steps: (1) 
installation of injection wells and associated equipment; (2) preparation of a dilute reagent 
solution from solid KMn 04 or a concentrated NaMn 04 solution; and (3) injection of the reagent 
solution. The reagent can be injected through the end of a direct push rod, through temporary 1- 
inch direct-push wells, or through permanent 2-inch or 4-inch conventionally-drilled wells. The 
selection of the most appropriate method for installing injection points depends on site-specific 
conditions including drilling costs, flow rate per well, and volume of fluid that must be injected. 

Permanganate can be distributed 5, 10, or 25 ft away from the injection point. However, 
achieving effective distribution often requires injecting large volumes of water. Depending on 
the injection well layout and formation permeability, water injection can require an hour to 
several days per well. As a consequence, several wells may be injected at one time using a 
simple injection system manifold. 

The primary design variables that must be considered when planning a Mn 04 injection project 
are: 

(1) spatial arrangement of the injection points; 

(2) type and physical construction of the injection points or wells; 

(3) amount of MnC >4 and water to inject; 

(4) reinjection frequency; and 

(5) additional labor and equipment required for mixing and injection. 

Each of these variables has an important influence on both the cost and effectiveness of the 
injection project. 

2.2.1 ARRANGEMENT OF INJECTION POINTS 

There are two general approaches used to distribute chemical reagents through the subsurface: 
(a) recirculation systems; and (b) injection only systems. 

Recirculation systems can be effective in distributing reagents significant distances through the 
subsurface in certain situations, allowing the use of fewer injection points. These systems are 
particularly useful where drilling costs are high or site access limitations restrict injection point 
installation. Recirculation systems can also be designed to minimize the physical displacement 
of contaminants by injection water. However, capital and operating costs of recirculation 
systems are often higher due to the more complex equipment and piping requirements and higher 
operation and maintenance (O&M) costs. In many cases, the design of recirculation systems is 
more complicated and may require the use of a site specific groundwater model. 

Injection only systems are most useful when drilling and site access conditions allow installation 
of rows or grids of injection points. Under these conditions, capital and O&M costs are often 
lower for injection only systems. The design of injection only systems can also be simplified by 
generating a ‘standard’ design for a small group of injection points which is then replicated 
throughout the site. The design tool described in this document has been developed to assist 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


4 


August 2010 



users in the design of injection only systems for distributing chemical reagents using grids of 
injection points or wells. 

Once the target treatment zone has been defined, the user must then select an injection point 
spacing. Selecting the best well spacing can be complicated. Increasing the separation between 
injection wells reduces the number of wells, reducing drilling costs. However, a larger well 
spacing can also increase the time required for injection, increasing labor costs. It may also be 
more difficult to uniformly distribute the reagent throughout the treatment zone using fewer, 
widely spaced injection points. In many cases, an intermediate well spacing results in the lowest 
total cost with reasonably good reagent distribution throughout the target treatment zone. The 
design tool provides output illustrating the effect of well spacing on distribution efficiency and 
comparative costs allowing the designer to select a well spacing that best meets project 
objectives. 

2.2.2 INJECTION POINT CONSTRUCTION 

Mn 04 solutions can be injected through the end of a direct push rod, through temporary 1-inch 
direct-push wells, or through permanent 2-inch or 4-inch conventionally-drilled wells. The 
selection of the most appropriate method for installing injection points depends on site-specific 
conditions including drilling costs, flow rate per well, and volume of fluid that must be injected. 

When the contamination extends over a significant vertical extent, it may be desirable to install 
several shorter screened wells to target specific intervals. This allows a known quantity of 
reagent to be injected in each interval. However, this also increases injection system cost and 
complexity. 

2.2.3 AMOUNT OF WATER AND CHEMICAL REAGENT TO INJECT 

Mn 04 is transported in the subsurface by flowing groundwater. Consequently, sufficient water 
must be injected to transport the Mn 04 throughout the target treatment zone. The amount of 
Mn 04 required is determined by the target treatment volume and the oxidant demand of the 
aquifer material. Mn 04 distribution in the aquifer can be enhanced by injecting more chemical 
reagent and/or more water. However, injecting additional reagent increases material costs and 
the potential for off-site migration of unreacted Mn 04 . Injecting additional water increases labor 
costs. The CDISCO design tool presented in Section 6 can be used to estimate the amount of 
reagent and water to inject. 

2.2.4 REINJECTION FREQUENCY 

Following injection, Mn 04 is consumed through reactions with the contaminant and Natural 
Oxidant Demand (NOD) associated with the aquifer material. To improve performance, 
additional Mn 04 may be injected. Section 5.4 presents information on the relative benefits of 
multiple injections on contaminant treatment. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


5 


August 2010 



2.2.5 ADDITIONAL LABOR AND EQUIPMENT REQUIRED 


The major capital costs for M 11 O 4 injection are associated with injection point installation, 
reagent purchase and labor during the injection. However, there are a number of other factors 
that can influence the final project cost including mobilization, setup of injection equipment (e.g. 
pumps, meters, etc.), injection water supply, and site cleanup. These costs are not closely related 
to the specific injection design. However, they can have a significant impact on the final project 
cost. In the design tool, costs for engineering and permitting, mobilization, equipment setup, 
water supply and cleanup/demobilization are entered as fixed costs. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


6 


August 2010 



3.0 PERMANGANATE CONSUMPTION BY AQUIFER MATERIAL 

3.1 INTRODUCTION 

When Mn 04 is injected into the subsurface, a portion of the material reacts with the target 
contaminant and a portion reacts with the aquifer material. The amount of permanganate that 
reacts with non-target chemicals is often referred to as the natural oxidant demand (NOD). 
When estimating the required amount of reagent to inject, designers must account for both 
permanganate consumed by the target chemicals and the NOD. If the NOD is not considered, 
the permanganate will be depleted more rapidly than expected and treatment efficiency may be 
less than desired. 

3.2 FATE AND TRANSPORT OF PERMANGANATE IN THE SUBSURFACE 

Permanganate has been used for wastewater and drinking water treatment for many years (Steel 
and McGhee 1979; Eilbeck and Mattock 1987). In contrast, permanganate has been used for 
groundwater remediation for less than 20 years. However, recent work has shown permanganate 
to be an effective oxidant for treatment of chlorinated solvents (perchloroethene [PCE] and 
trichloroethene [TCE]) and aromatic hydrocarbons (naphthalene, phenanthrene, pyrene and 
phenols) (Vella et al. 1990; Leung et al. 1992; Vella and Veronda 1992; Gates et al. 1995, 2001; 
Yan and Schwartz 1996; 1999; Schnarr et al. 1998; West et al. 1998; Siegrist et al. 1998a, b, 
2000, 2001; Lowe et al. 1999). 

Representative reactions illustrating the oxidation of PCE (C2CI4) and TCE (C2HCI3) by 
permanganate are shown below. 

C 2 CI 4 + 1V 3 M 11 O 4 " 1V 3 Mn0 2 (s) + 2C0 2 + 2 2 / 3 H + + 4Cf (PCE) 

C 2 HC1 3 + 2Mn0 4 ' 2Mn0 2 (s) + 2C0 2 + H + + 3C1' (TCE) 

Based on this stochiometry, 1.81 g of permanganate is needed to degrade 1 g of TCE, releasing 
1.32 g of manganese dioxide, 0.67 g of carbon dioxide and 0.81 g of Cf (Siegrist et al. 2001). 

3.2.1 NATURAL OXIDANT DEMAND (NOD) 

In many cases, the NOD controls the amount of reagent which must be injected for effective 
treatment (Marvin et al. 2002). NOD is exerted when permanganate reacts with a variety of 
naturally occurring materials including ferrous iron, sulfides and natural organic carbon. NOD is 
commonly measured by reacting aquifer material with a permanganate solution and measuring 
change in permanganate concentration over time. NOD is typically reported as mass of 
permanganate consumed per unit mass of aquifer solids (Siegrist et al. 2000; Marvin et al. 2002). 
Efforts are underway to standardize the NOD test procedure (ASTM 2007). However, this 
standard protocol has only been used by a few investigators and published values of NOD have 
been measured under a range of experimental conditions. Table 3.1 shows the measured values 
of NOD value for several different sites and experimental conditions. Reported NOD values 
range from 0.3 to 88 g Mn 04 / Kg indicating NOD can vary dramatically between sites. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


7 


August 2010 



Table 3.1: Reported Values of NOD for Different Sites. 


Source 

NOD 

Site 

Description 

g/Kg 

mmol/Kg 

Mumford et al., 
2005 

1.2 

10 

CFBase Borden 

Batch test 

0.35 

3 

Column test 

Mumford et al., 
2004 

0.51-0.75 

4.3-6.3 

CFBase Borden 

Push-pull test Static 

0.29-0.42 

2.4-3.5 

Push-pull test, 
Drift or Reaction 

Urynowicz, 2008 

2.9 

24 

- 

Low dose, 48 hr 

4.8 

40 

- 

Mid dose, 48 hr 

6.1 

51 

- 

High dose, 48 hr 

23.2 

195 

- 

High dose, 6 weeks 

Oberle et al., 
2000 

0.2 

1.68 

Northern Ohio 

Saturated Sand 

15-20 

126-168 

From MI 

Unsaturated Sandy 
Clay 

Chambers et al., 
2000 

7.61 

64 

From CA 

Silt, 14 days 

7.16 

60 

Clay, 14 days 

4.49 

38 

Sand, 14 days 

Drescher et al. 1998 

30 

252 

- 

- 

Siegrist et al. 2001 

11 

92 

- 

- 

Hood 2000a 

1 

8 

CFBase Borden 

- 

Xu 2006 

2.12 

17.8 

CFB Borden 

fine/medium sand 

2.28 

19.2 

National Test Site, 
Dover AFB, DE 

sandy loam 

32.29 

27.1 

East Gate Disposal Yard, 
Fort, Lewis, WA 

loamy sand, gravel, 
cobbles 

0.77 

6.5 

Milan Army 
Ammunition Plant, TN 

Sand 

11.42 

96.0 

Launch Complex 34, Cape 
Canaveral AFS, FL 

loamy 

coarse/medium sand 

5.5 

46.2 

87.9 

738 

NFF, Cecil Field, FL 

loamy fine sand 

2.54 

21.3 

NIROP, 

Bacchus Works Facility, UT 

gravels, loam 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


August 2010 



3.2.2 MODELING APPROACHES 


A number of different modeling approaches have been applied to describe the kinetics of Mn 04 
consumption by NOD (Hood and Thomson, 2000; Reitsma and Dai, 2001; Mumford, 2002; Xu, 
2006; Jones, 2007; Honning et al., 2007; Urynowicz et al., 2008). Potentially the simplest 
approach would be to assume that the reaction is instantaneous and NOD must be completely 
consumed before the Mn 04 can be transported away from the injection point. However, this 
approach would likely over-estimate Mn 04 consumption since a portion of the NOD may react 
slowly. The most common modeling approach has been to simulate the reaction between M and 
NOD (N) as a 2 nd order reaction (Zhang and Schwartz 2000, Xu 2006), 

dM/dt = - k N M N 

where kN is the 2 nd order rate of reaction between M 11 O 4 and NOD. Zhang and Schwartz 2000 
used a kN value of 450 M ' 1 s ’ 1 which is much faster than the rate of contaminant oxidation and 
results in essentially complete depletion of the NOD before Mn 04 will migrate through the 
aquifer. However in batch experiments conducted by Honning et al. (2007), the long-term 
consumption of Mn 04 by NOD could not be described by a single 1 st order decay rate. During 
the first few hours of the reaction, Mn 04 decreased at rates of 0.05 - 0.5 h ' 1 and then declined 
more slowly. 

Recent work has shown that NOD is composed of several components or fractions with varying 
reactivity (Mumford et al., 2005; Xu, 2006; Urynowicz et al., 2008). Ideally, NOD would be 
represented with a continuum of reaction rates where the less reactive fraction becomes 
progressively more important as the more reactive NOD fraction is depleted. However, modeling 
studies by Xu (2006) and Urynowicz et al. (2008) suggest that Mn 04 consumption by NOD can 
be reasonably well described assuming the NOD is composed of one fast and one slow fraction. 
In batch experiments conducted by Xu (2006), a fraction of the NOD was depleted in a few 
hours followed by much slower degradation, where the slow NOD reaction rate varied from 
0.014 to 0.72 L/mmol-day with a median value of 0.077 L/mmol-day. In batch experiments 
conducted by Urynowicz et al. (2008), the fast NOD appeared to be consumed with about 48 
hours, followed by slower depletion of Mn 04 at rates of 0.024 to 0.13 d' 1 , depending Mn 04 dose. 

3.3 KINETIC MODEL EVALUATION 

At present, there is no general consensus on the best approach for simulating Mn 04 consumption 
by NOD. However, there does seem to be some agreement that: (1) NOD is often composed of 
different components or fractions; ( 2 ) some components react fairly quickly (minutes to hours); 
(3) some components react more slowly (days to months); and (4) the effective NOD is a 
function of permanganate concentration with higher concentrations resulting in higher effective 
NOD. 

In this project, groundwater flow, transport and reaction models (MODFLOW and RT3D) are 
used to evaluate the effect of injection conditions on treatment efficiency in a three dimensional 
(3D) heterogeneous aquifer. The models must adequately represent the kinetics of Mn 04 
consumption by NOD, but must be relatively simple to implement and not result in an excessive 

ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 9 August 2010 



computational burden. Given that there is no one modeling approach that is generally accepted, 
six different approaches were evaluated for describing NOD kinetics. Each of the modeling 
approaches was calibrated to match the results of all batch NOD incubations with a sample of 
aquifer material (Soil C) from the Massachusetts Military Reservation (MMR). Experimental 
data was provided by Dr. Michelle Crimi of Clarkson University. The model performance was 
then evaluated based on a visual comparison of measured and simulated NOD, and statistical 
measures describing the goodness of fit. The batch incubations were conducted in glass jars with 
varying amounts of aquifer material and KMn 04 (500, 1000 and 5000 mg/L), aquifer material 
(16, 32 and 48 g) and water (10, 20 and 30 g) (see Table 3.2 for the different treatments). In 
addition, three different mixing conditions were evaluated: (1) complete mixing; (2) mixing once 
per day; and (3) static (mix only before sample collection). MnC >4 concentration was typically 
measured approximately 16 times over the 28 day incubation period resulting in a total of 834 
MnC >4 measurements. 


Table 3.2: Batch Experimental Conditions of Each Treatment. 


Treatment 

Initial KMnC >4 Cone. (mg/L) 

Mass Solids (g) 

Mass Water (g) 

1 

5000 

16 

30 

2 

1000 

16 

30 

3 

500 

16 

30 

4 

5000 

32 

20 

5 

1000 

32 

20 

6 

500 

32 

20 

7 

5000 

48 

10 

8 

1000 

48 

10 

9 

500 

48 

10 


The kinetic models evaluated in this work are summarized below. Each model was coded into 
MS Excel as a Visual BASIC subroutine using a 4th order Runge-Kutta solution of the ordinary 
differential equations (ODE’s). An initial set of model coefficients was assumed, and used to 
predict the variation in MnC >4 concentrations vs. time for all incubations of Soil C. The goodness 
of fit was then evaluated as the root mean square error (RMSE) between simulated and measured 
MnC >4 concentration in all incubations of Soil C. Best fit parameter values were found using the 
Solver function in Excel to search for the parameter set that resulted in the smallest RMSE. Best 
fit parameter values were obtained for the three soils for each mixing condition (complete mix, 
mix once per day, and static) and a fourth condition where all the data for the soil was pooled 
together, ignoring mixing condition. 

For each of the models, the ME (Mean Error) and the RSME are presented as indicators of 
goodness of fit. The best model should have a value of ME and RMSE close to zero. Graphs are 
also presented showing the measured change in MnC >4 concentration from the start of the 
incubation (AMn 04 ) vs. simulated AMnC >4 with the pooled data for Soil C. Ideally, the 
experimental measurements should cluster around the 45° line indicating a 1:1 match between 
measured and simulated values. Clustering of data away from the 45° line indicates that there is 
some trend in the experimental results that is not captured by the kinetic model. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


10 


August 2010 



In each of the models presented below, M is the Mn0 4 concentration (mol L' 1 ), Pb is the soil bulk 
density (Kg L' 1 ) and n is porosity (dimensionless). 

3.3.1 MODEL 1 - ZERO ORDER LOSS OF MNO 4 

Model 1 assumes that Mn0 4 concentration declines at a constant (zero order) rate with time, 
independent of the amount of NOD. This very simplified approach was included to provide a 
comparison with more complex kinetic representations. The change in permanganate 
concentration (M) is represented by the equation 3-1. 

dM 

~T = ~ k ° 

dt (3-1) 

where, 

k 0 = zero order Mn0 4 consumption rate (mmol L' 1 d' 1 ) 

Experimental results are compared to simulated values for Model 1 in Figure 3.1 and Table 3.3. 
As expected, Model 1 provided a relatively poor fit to the experimental data. Visual examination 
of the results shown in Figure 3.1 show that simulated values of AMn0 4 are clustered near the x- 
axis and do not increase with increasing values of measured AMn0 4 . 


Table 3.1: Statistical Results of Model 1 Evaluation. 


Mixing Condition 

k 0 

Mean 

Error 

RMSE 


mmol/L- 

day 

mmol/L 

mmol/L 

Complete condition 

0.188 

-0.545 

1.582 

Once per day condition 

0.202 

-0.470 

1.747 

Static condition 

0.166 

-0.274 

1.139 

Total condition 

0.185 

-0.430 

1.513 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


11 


August 2010 




Experimental AMnD4 [limioLL] 


Figure 3.1: Comparison of Observed Values of AMn 04 with Model 1 Simulation Results 

(all data for Soil C). 

3.3.2 MODEL 2 - FIRST ORDER LOSS OF MN0 4 

Model 2 assumes that M 11 O 4 concentration declines at a first order rate with time, independent of 
the amount of NOD. The change in permanganate concentration (M) is represented by the 
equation 3-2. 


dt 


— k x U M 


where, 


(3-2) 


k IM = 1 st order Mn0 4 consumption rate (d 1 ) 


Experimental results are compared to simulated values for Model 2 in Figure 3.2 and Table 3.4. 
Model 2 provided a slightly better fit to the data than Model 1. However, the fit was still 
relatively poor. One difference is that Model 2 predicted higher values of AMn0 4 for higher 
initial Mn0 4 concentrations. This resulted in a small improvement in the goodness of fit 
statistics. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


12 


August 2010 
















Table 3.4: Statistical Results of Model 2 Evaluation. 


Mixing Condition 

klM 

Mean Error 

RMSE 


1/day 

mmol/L 

mmol/L 

Complete condition 

0.014 

-0.587 

1.472 

Once per day condition 

0.017 

-0.458 

1.530 

Static condition 

0.013 

-0.291 

0.955 

Total condition 

0.015 

-0.446 

1.349 



Experimental AMnC4 [inmol/L] 


Figure 3.2: Comparison of Observed Values of AMn 04 with Model 2 Simulation Results 

(all data for Soil C). 

3.3.3 MODEL 3 - FIRST ORDER LOSS OF NOD 


Model 3 assumes that NOD concentration declines at a first order rate with time, and the change 
in Mn 04 concentration is equal to the change in NOD. This very simple approach was included 
to provide a comparison with more complex kinetic representations. The change in 
permanganate concentration (M) is represented by the equations 3-3a and 3-3b. 


dN 


dt 


F _ . 


-K n n 


dM_ 

dt 


PbKn^ / n 


(3-3 a) 
(3-3b) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


13 


August 2010 


























where, 


k IN = 1 st order NOD consumption rate (d 1 ) 

Experimental results are compared to simulated values for Model 3 in Figure 3.3 and Table 3.5. 
Model 3 provided a significantly better fit to the data than Models 1 and 2. However, visual 
examination of Figure 3.3 shows that the fit is still less than desired. There appears to be clusters 
of data above and below the 45° line, indicating there is one group of measurements where 
Model 3 consistently under predicts the observed AMnOzj, and a second group of data where 
while Model 3 consistently over predicts the observed AMnCV 


Table 3.5: Statistical Results of Model 3 Evaluation. 


Mixing Condition 

km 

NOD initial 

Mean Error 

RMSE 


1/day 

mmol/g 

mmol/L 

mmol/L 

Complete condition 

1.920 

0.00079 

-0.182 

1.237 

Once per day condition 

0.807 

0.00096 

-0.155 

1.412 

Static condition 

0.316 

0.00084 

-0.189 

1.021 

Total condition 

0.791 

0.00083 

-0.189 

1.266 



Experimental AMn04 [imiiol/L] 


Figure 3.3: Comparison of Observed Values of AM 11 O 4 with Model 3 Simulation Results 

(all data for Soil C). 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


14 


August 2010 


























3.3.4 MODEL 4 - SECOND ORDER LOSS OF MNO 4 AND NOD 


Model 4 assumes that the concentrations of Mn0 4 and NOD both decline as a first order function 
of both Mn0 4 and NOD (second order function overall). This approach has been used by a 
number of investigators including Zhang and Schwartz (2000) and Xu (2006). Changes in 
permanganate concentration (M) and NOD concentration (N) are represented by the equations 3- 
4a and 3-4b. 

dM 

-= -p B k 2 NM / n 

dt 

dN , Ar)1 ^ 

-= -kjNM 

dt 2 

where, 

k 2 = 2 nd order NOD consumption rate (L mmol' 1 d' 1 ) 

Experimental results are compared to simulated values for Model 4 in Figure 3.4 and Table 3.6. 
Model 4 provided a significantly lower ME and RMSE compared to Model 3 indicating a 
significant improvement in model fit. However, there is no obvious clustering of the data, 
indicating there are no consistent trends that are not captured by the model. 


(3-4a) 
(3-4b) 


Table 3.6: Statistical Resul 

Its of Model 4 

1 Evaluation. 

Mixing Condition 

k 2 

NOD 

initial 

Mean 

Error 

RMSE 


L/mmol- 

day 

mmol/g 

mmol/L 

mmol/L 

Complete condition 

0.046 

0.00174 

-0.070 

0.649 

Once per day condition 

0.022 

0.00236 

-0.033 

0.629 

Static condition 

0.011 

0.00210 

-0.118 

0.508 

Total condition 

0.022 

0.00201 

-0.092 

0.686 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


15 


August 2010 




Experimental AMnC)4 [lmnol/L] 

Figure 3.4: Comparison of Observed Values of AMn0 4 with Model 4 Simulation Results 

(all data for Soil C). 

3.3.5 MODEL 5 - SECOND ORDER LOSS OF MN0 4 WITH FAST AND SLOW NOD 


Model 5 is similar to Model 4. However, the NOD is assumed to be composed of a fast fraction 
and a slow fraction, similar to the modeling approach employed by Jones (2007) and Urynowicz 
et al. (2008). The change in permanganate (M), fast NOD (Nf), and slow NOD (Ns) are 
represented by equations 3-5a, b and c. 


dM_ 

dt 


-p B k 2F N F MIn-p B k 2S N s M /n 


dN F 
dt 


- -k 1F N F M 


dN s _ 

dt 


— k 2S N s M 


(3-5a) 

(3-5b) 

(3-5c) 


where, 

k 2F = 2 nd order Fast NOD consumption rate (L mmol" 1 d" 1 ) 
k 2S = 2 nd order Slow NOD consumption rate (L mmol" 1 d" 1 ) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


16 


August 2010 


















Experimental results are compared to simulated values for Model 5 in Figure 3.5 and Table 3.7. 
RMSE values for Model 5 were very similar to Model 4 indicating that separation of the NOD 
into a fast and slow fraction did not significantly improve the model fit for the MMR soils. 
However, ME values were much lower. In addition, Model 5 run times were significantly longer 
than for Model 4. This was because the high reaction rates for fast NOD required a shorter 
computational time step. 


Table 3.7: Statistical Results of Model 5 Evaluation. 


Mixing Condition 

k2F 

k2S 

Fraction 

Fast 

NOD 

initial 

Mean Error 

RMSE 


L/mmol-d 

L/mmol-d 


mmol/g 

mmol/L 

mmol/L 

Complete condition 

0.0458 

0.000001 

0.15 

0.0117 

0.017 

0.650 

Once per day 
condition 

0.0223 

0.000001 

0.14 

0.0170 

0.014 

0.629 

Static condition 

0.0107 

0.000005 

0.16 

0.0128 

-0.015 

0.508 

Total condition 

0.0220 

0.000001 

0.15 

0.0132 

-0.001 

0.686 


Figure 3.5: 


eJ 

o 

1 

Q 

3 

]S 

I 



Experimental ANJ11O4 


Comparison of Observed Values of AM 11 O 4 with Model 5 Simulation Results 

(all data for Soil C). 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


17 


August 2010 



























3.3.6 MODEL 6 - SECOND ORDER LOSS OF MNO 4 WITH INSTANTANEOUS AND 
SLOW NOD 

Model 6 is similar to Model 5. However, the reaction between Mn0 4 and fast NOD is assumed 
to occur so quickly, that it is essentially instantaneous. The instantaneous change in 
permanganate (M) and instantaneous NOD (Ni) are calculated by an if->then statement, 

When concentration of M > Nj* p g / n 

M = M - p B * Nj / n and Nj =0 otherwise M = 0 and N t = N, —M*n/p B 

Once the instantaneous reaction is complete, the change in permanganate (M) and slow NOD 
(Ns) are calculated by solving equations 3-6a and 3-6b. 

dM , Ar ., . 

—— — -p B k 2S N s M / n 
dt 

= N M 

dt 2S s 

k 2S = 2 nd order Slow NOD consumption rate (L mmol' 1 d' 1 ) 

Experimental results are compared to simulated values for Model 6 in Figure 3.6 and Table 3.8. 
Error statistics for Model 6 were very similar to Models 4 and 5 indicating that representing the 
reaction between ‘fast’ NOD and Mn0 4 as an instantaneous reaction did not significantly hurt 
model performance. However, Model 6 run times were significantly shorter than Model 5. This 
would be a major advantage when simulating complex 3D aquifers. 


Table 3.8: Statistical Results of Model 6 Evaluation 


Mixing Condition 

k2S 

Fraction 

Instantaneous 

NOD 

initial 

Mean 

Error 

RMSE 


L/mmol-day 


mmol/g 

mmol/L 

mmol/L 

Complete condition 

0.0457 

0.001 

0.0017 

-0.069 

0.649 

Once per day 
condition 

0.0221 

0.001 

0.0024 

-0.031 

0.630 

Static condition 

0.0106 

0.001 

0.0021 

-0.117 

0.508 

Total condition 

0.0217 

0.002 

0.0020 

-0.088 

0.685 


(3-6a) 

(3-6b) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


18 


August 2010 



12.00 


Figure 3.6: 


«=? 


o 

i 



10.00 


8.00 


6.00 


4 00 


2.00 


0.00 

-2.00 0.00 2.00 4.00 6.00 8.00 10.00 12.00 

Experimental AN Jii04 

Comparison of Observed Values of AMn 04 with Model 6 Simulation Results 

(all data for Soil C). 



3.3.7 KINETIC MODEL EVALUATION SUMMARY 


Six kinetic models were developed and calibrated to experimental measurements of NOD 
exerted by Soil C from MMR. The zero order and first order models (Models 1, 2 and 3) 
provided a relatively poor fit to the experimental results and were not be considered further. 

All of the 2 nd order models (4, 5 and 6) provided a relatively good fit to the experimental results. 
The RMSE was 0.54 for all 2 nd order models using all of the Soil C data (total condition). The 
essentially identical performance of Models 4, 5 and 6 should not be surprising given the similar 
structure of each model. 


Table 3.9 compares the estimated coefficients for Models 4, 5 and 6. In Model 5, the slow NOD 
reaction coefficient is so low (less than 10' 6 L/mmol-d), that the slow NOD is essentially non¬ 
reactive over the 28 day incubation period. Consequently, the estimated values of total NOD in 
Model 4 are identical to fast NOD values in Model 5 and 2 nd order rate coefficients in Model 4 
are essentially the same as the fast 2 nd order rate coefficients in Model 5. For practical purposes, 
the best fit parameter estimates for Models 4 and 5 are identical, and consequently the 
performance of these models is equivalent. Similarly, estimated values of total NOD for Models 
4 and 6 are identical and performance of these models in matching the experimental data is 
equivalent. Estimated values for the 2 nd order reaction rate in Model 6 are slightly lower than for 
Model 4 because of a small portion of the NOD in Model 6 (1-2%) reacts instantaneously. In 
summary, the overall performance and parameter estimates generated by Models 4, 5 and 6 are 
essentially identical for MMR Soil C. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


19 


August 2010 














Table3.9: Best! 

Fit Coefficients for 

Model 4, 5, and 6. 

Mixing 

Condition 

Model 4 

Model 5 

Model 6 

NOD 

initial 

k 2 

NOD f 

k 2 F 

k 2 s 

Total 

NOD 

Instant 

Fraction 

k 2 s 


mmol/g 

L/mmol-d 

mmol/g 

L/mmol-d 

L/mmol-d 

mmol/g 


L/mmol-d 

Complete 

condition 

0.00174 

0.046 

0.0117 

0.0458 

0.000001 

0.0017 

0.001 

0.0457 

Once per day 
condition 

0.00236 

0.022 

0.017 

0.0223 

0.000001 

0.0024 

0.001 

0.0221 

Static condition 

0.00210 

0.011 

0.0128 

0.0107 

0.000005 

0.0021 

0.001 

0.0106 

Total condition 

0.00201 

0.022 

0.0132 

0.022 

0.000001 

0.002 

0.002 

0.0217 


Given that Models 4, 5, and 6 all provide an equally good fit to the experimental data and 
provide similar parameter estimates, there is no specific reason to select one model over another. 
However, prior research by other investigators (Mumford et al. 2005, Mumford et al. 2004, Xu 
2006, and Urynowicz et al., 2008) indicates that soil NOD often contains fractions that react 
more rapidly and more slowly. Models 5 and 6 have the capability of simulating fast and slow 
fractions, while Model 4 does not. Model 5 did not provide a significantly improved fit 
compared to Model 6, but did take significantly longer to run due to the short computational time 
step required. Model 6 was used in future modeling work. Model 6 performed as well as Model 
4, retains the capability of simulating fast and slow fractions, and did not require significantly 
longer run times than Model 4. If the instantaneous fraction in Model 6 is set to zero, Models 4 
and 6 are identical. 

3.4 MMR PARAMETER ESTIMATES 

In Section 3, a field scale groundwater flow, transport and reaction model were be used to 
simulate MnC >4 distribution and contaminant destruction in a large scale pilot test conducted at 
MMR. Estimates of the Model 6 kinetic parameters were required to calibrate this model. 

Soil samples were collected from three depths (Samples A, B and C) in the vicinity of the MMR 
pilot test and tested by Dr. Michelle Crimi to provide NOD estimates. Characteristics of these 
soils are summarized in Table 3.10. Soil B was visually different than Soils A and C with less 
silt. Soil A is collected from the deepest location, C is middle and B is the shallowest location of 
injection well. The Unified Soil Classification System (USCS) define Soil A as SM and Soil B 
as SP which means silty sand and poor graded sand as well. Soil C is mix of undefined soil and 
SM. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


20 


August 2010 



Table 3.10: MMR Soil Sample Comparison. 


Soil 

A 

B 

C 

Depth (ft) 

185’ -189’ 

157’- 159.3’ 

172’-176.3’ 

uses 

SM 

SP 

SM + undefined 

Texture 

Silt and Sand 

Sand 

Silt and unknown soil 

Hydro. Conductivity (ft/day) 

22 

194 

22 to 194 


The experimental protocol followed for the three soils samples were similar, but not identical. 
Soil C was tested with all three mixing treatments (complete mixing, mix once per day, and 
static). Soil A was tested with static and once per day mixing. Soil B was tested with complete 
and once per day mixing. This resulted in a total of ten mixing-soil combinations resulting in ten 
sets of parameter estimates. Model 6 kinetic parameters were estimated for each soil-mixing 
combination following the procedure previously described. Estimated values of total NOD, the 
slow NOD reaction rate, and fraction instantaneous NOD are shown in Table 3.11 for each of the 
ten soil-mixing combinations. 


Table 3.11a: Best Fit Parameter 


Mixing Condition 


Static 


Once per day 


Complete 


All Data 


Estimates for MMR Soils 


Soil A 


0.0026 


0.0026 


0.0026 


Soil B 


0.0054 


0.0064 


0.0054 


otal NOD (mm ol/g) 


Soil C 


0.0036 


0.0036 


0.0036 


0.0036 


Table 3.11b: Best Fit Parameter Estimates for MMR Soils - Slow Reaction Rate (k2S) 
_ (L/mmol-d) __ 


Mixing Condition 

Soil A 

Soil B 

Soil C 

Static 

0.006 


0.003 

Once per day 

0.007 

0.008 

0.005 

Complete 


0.018 

0.007 

All Data 

0.007 

0.016 

0.004 


Table 3.11c: Best Fit Parameter Estimates for MMR Soils - Fraction Instantaneous 


Mixing Condition 

Soil A 

Soil B 

Soil C 

Static 

0.070 

NA 

0.070 

Once per day 

0.111 

0.107 

0.111 

Complete 

NA 

0.044 

0.019 

All Data 

0.090 

0.060 

0.090 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


21 


August 2010 





The total NOD of Soil B was substantially higher than for Soils A and C, which is consistent 
with the different visual appearance of this material. The instantaneous fraction was less than 
11% for all the soils varying from 6.4% tolO.5% of the total NOD. The Slow Reaction Rate (£?,s) 
for all soils and mixing conditions was relatively consistent varying from 0.003 to 0.022 
L/mmol-d. An initial value of the effective 1 st order decay coefficient (Ke) for permanganate can 
be estimated as 


Ke = p B k 2S N s / n 


Using this approach, the initial effective decay rate for permanganate varied from 0.1 to 1.0 per 
day with an average of 0.5 per day. This indicates the ‘slow’ NOD is reasonably reactive and 
could result in reasonably rapid depletion of permanganate. 

In Section 3, a model is developed and applied to simulate the ISCO of TCE in a large scale field 
pilot test at MMR. Estimates of each of the Model 6 kinetic parameters were required to 
calibrate this model to the MMR field site. The results presented above indicate that mixing 
condition has a modest impact of the different kinetic parameters. As a consequence, all the data 
for the different mixing conditions was grouped together and analyzed as a single data set. Table 
3.12 shows the best estimates of the Model 6 parameters for three groups of data: (1) Soil A and 
C combined, (2) Soil B and (3) Soil A, B and C combined. Soils A and B were grouped together 
based on their similar NOD characteristics and visual appearance. 


Table 3.12: Parameter Set for MMR 


Soil 

Type 

Fraction 

Instantaneous 

Initial 

NOD 

Slow Reaction Rate 

N* 

ME 

RMSE 



mmol/g 

L/mmol-d 


mmol/L 

mmol/L 

A,C 

0.0903 

0.0031 

0.0051 

57 

0.0047 

0.692 

B 

0.0597 

0.0054 

0.0163 

12 

-0.1539 

1.031 

A,B,C 

0.0801 

0.0039 

0.0089 

69 

-0.04816 

0.805 


* N is the number of individual incubations used to generate these estimates. Each incubation 
typically had 16 separate permanganate measurements. 


3.5 SUMMARY AND CONCLUSIONS - MN0 4 CONSUMPTION BY AQUIFER 
MATERIAL 


Permanganate (Mn0 4 ) injection can be a very effective technology for in situ treatment of certain 
chlorinated solvents. However, effective Mn0 4 distribution is often controlled by the NOD of 
the aquifer material and groundwater. At present, there is no general consensus on the best 
approach for simulating Mn0 4 consumption by NOD. However, there does seem to be some 
agreement that: (1) NOD is often composed of different components or fractions; (2) some 
components react fairly quickly (minutes to hours); (3) some components react more slowly 
(days to months); and (4) the effective NOD is a function of permanganate concentration with 
higher concentrations resulting in higher effective NOD. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


22 


August 2010 



Six different kinetic relationships were examined to identify the relationship that best fits the loss 
of permanganate in experimental incubations using soil C from MMR. The three models that 
included a second order relationship between permanganate concentration (M) and NOD 
concentration all provided an equally good fit to the experimental data. Model 6 was selected for 
future use based on its’ ability to fit the experimental data, ease of numerical solution, and 
flexibility in simulating NOD composed of rapidly and slowly reactive materials. Model 6 
includes a fraction of NOD which instantaneously reacts with Mn 04 and a slow NOD component 
which reacts with Mn 04 by a second order relationship. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


23 


August 2010 



4.0 MODEL TESTING - MMRISCO PILOT TEST 


4.1 INTRODUCTION 

A pilot test of ISCO using permanganate was conducted at the Massachusetts Military 
Reservation in Fall 2007 to evaluate the potential applicability of ISCO for treatment of the CS- 
10 contaminant plume. Results from this pilot test were be used to evaluate the capability of a 
groundwater flow, transport and reaction model to simulate ISCO in the subsurface. This model 
was developed using the previously described Model 6 and NOD kinetic parameters estimated in 
Section 2. Three different parameter sets (Table 3.12) which were generated from laboratory 
NOD measurements and are used in the model calibration. 

4.2 MASSACHUSETTS MILITARY RESERVATION (MMR) 

Massachusetts Military Reservation (MMR) is located on the south-western portion of Cape 
Cod, Massachusetts near the towns of Bourne, Mashpee, Sandwich, and Falmouth (Figure 4.1). 



The MMR has been used for military purposes since 1911. Although the occupants and 
boundary have been changed since MMR was established in the 1930s, the facility has always 
provided training and housing space for Air Force and/or Army units. Work at MMR includes 
training and maneuvers, military aircraft/vehicle operations, maintenance (repair, cleaning, oil 
change and body work), and support. The hazardous wastes generated at the site were 
commonly disposed in landfills, drywells, sumps, and occasionally burned at firefighter-training 
areas. This has resulted in multiple contaminant plumes. The contaminants detected in 
groundwater at MMR include carbon tetrachloride (CCI4), TCE, PCE, ethylene dibromide 
(EDB), and benzene. Hydrologic condition at MMR result in high groundwater flow velocities 
(1-2 feet per day) and a radial flow pattern where the plumes migrate outward from the center of 
the site towards water bodies on the site boundaries (Figure 4.2) (AFCEE 2007a). 

ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 24 August 2010 





Chemical Spill 10 (CS-10) located in the southern portion of MMR is the focus of this study. 
Contaminants of concern in the CS-10 plume are TCE and PCE with maximum observed 
concentrations of 602 and 119 pg/L, respectively. These values are above the allowable 
Maximum Contaminant Levels (MCL) of 5 pg/L for both TCE and PCE (AFCEE 2007b). A 
pump and treat facility (CS-10 Treatment Facility) is operated to control migration of these 
contaminants. The pilot study described below was conducted to determine if ISCO would be 
effective in reducing contaminant concentrations in the CS-10 plume and potentially reducing 
the required operating period of the CS-10 Treatment Facility. 



Ma$»ACh|i$«||$ Milifjiry 
Reservation / 


C&-1QTr*airt»M Fadll* 


CS-10 


Stuidwnb 
Tnsairwnt Factftv 


A^Htv/ 




ruwnaf , 

Sandwich 






{ Tfflri uf ^ rig __ 

Bourne^ / 


\a 


Tl ■ III r >1 


r ■" ttrC 


FS-1 Triialmird 
Facikty 




FS-Z0 TfWMIWf FflCJily 


ImTuif 


Falmouth \ 






Figure 4.2: Plume Distribution of MMR (grey area represent MMR, red line represent 


plume boundary, AFCEE 2007b). 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


25 


August 2010 



































4.3 PILOT TEST 



Massacliusetts Military 
Reservation J 


Sojths"! 
Trench Area 


r Falmouth 


tJortti-Centra I 


The ISCO pilot test was conducted near monitor well 03MW1024 because of the relatively high 
TCE concentrations in this area and the opportunity to reduce future remediation costs. The 
geology of this site is similar to other portions of the CS-10 plume and other areas at MMR, 
potentially allowing results from this site to be extrapolated to other locations. The test area is 
also easily accessible for power supplies, equipment, personnel and is a secure location (CH2M 
Hill 2007). 


Figure 4.3: CS-10 Plume (grey area represents MMR, red line represent plume boundary, 

AFCEE 2007b). 


TCE contamination at the pilot test site is distributed in two major zones (Figure 4.4), a shallow 
zone extending from 150 to 205 feet below ground surface (ft bgs) with concentration ranging 
from TCE varying from 18 to 590 pg/L, and a deep zone extending from 230 to 295 ft bgs with 
TCE varying from 12 to 98 pg/L. Most of the aquifer material in pilot test area is sand. 
However, a fine sand /silty sand unit is located between 175 and 201 ft bgs, at approximately the 
same depth as the zone of maximum contaminant concentration (CH2M Hill 2007). 


CS-10 In-Plume Treatment Plant 


Bourne 


CS-10 HvPTume 
Treatment Faculty 


A 

r\ 

■ Pluffl« Flows « 

A 


U*d?r Rand p 

* 

^ \> 


1.300 

2JS» 



^Mashpec; 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


26 


August 2010 




















Figure 4.4: Cross Section of Pilot Test Area (CH2M Hill 2007). 



ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


27 


August 2010 







































The pilot test area contained a single injection well and seven monitoring wells. The injection 
well (03IJW001) had three separate screened intervals allowing injection at different depths. 
The monitoring wells were installed up-gradient, cross-gradient, and down-gradient of the 
injection location with 22 well screens allowing monitoring at different depths. Table 4.1 
provides details on the screened interval of each injection and monitoring well. 

The pilot test design consisted of injecting approximately 5,000 mg/L NaMnCE for 10 hours per 
day for four days. This solution was to be injected into three injection wells simultaneously at 10 
gallons per minute (gpm) for a total flow of 30 gpm. The plan was to inject a total volume of 
72,000 gallons of water containing 1363 Kg NaMnOzj. The actual injection flow rates and 
concentrations varied somewhat from this plan due to problems in maintaining a constant 
dilution rate and difficulty in maintaining high injection rates in the lower permeability zones. 

TCE and permanganate concentrations in the injection and selected monitor wells were measured 
five times between November 2007 and May 2008. Monitoring results are presented in Tables 
4.2 and 4.3. Permanganate concentrations were determined by measuring manganese 
concentration in the water with a Hach Test Kit. TCE was not monitored when the groundwater 
had a purple color indicating the presence of dissolved permanganate. For model calibration 
purposes, the TCE concentration was assumed to be zero whenever dissolved permanganate was 
present as indicated by a purple color. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


28 


August 2010 



Table 4.1: Well Construction Information. 


Well Cluster 

Purpose 

(target of monitoring) 

General Location 
(from injection well) 

Top of Screen 
(ft bgs) 

Bottom of Screen 
(ft bgs) 

03IJW001A 

Oxidant injection 

- 

185.3 

190.3 

03IJW001B 

172.2 

177.2 

03IJW001C 

157.15 

162.15 

03MW1024A 

Vertical migration of 
the injected oxidant 

25 ft 

Down-gradient 

245.29 

250 

03MW1024B 

210.47 

215 

03MW1024C 

200.29 

205 

03MW1024D 

190.29 

195 

03MW1024E 

180.02 

185 

03MW1024F 

162.02 

167 

03MW1024G 

152.24 

157 

03MW1024H 

142.24 

147 

03MW1032A 

TCE concentrations 
approaching the test 
area 

30 ft 

Up-gradient 

185.37 

190.37 

03MW1032B 

172.63 

177.63 

03MW1032C 

157.46 

162.46 

03MW1033A 

Tracking the changes 
in oxidant and TCE 
along the centerline of 
the flow 

60 ft 

Down-gradient 

200.1 

205.1 

03MW1033B 

172.28 

177.28 

03MW1033C 

147.32 

152.32 

03MW1034A 

Horizontal migration 
of the injected 
permanganate 

35 ft 

SE Down-gradient 

200.38 

205.38 

03MW1034B 

173.37 

178.37 

03MW1034C 

147.08 

152.08 

03MW1035A 

35 ft 

SW Down-gradient 

200.02 

205.02 

03MW1035B 

172.41 

177.41 

03MW1035C 

152.04 

157.04 

03MW1036A 

65 ft 

SE Down-gradient 



03MW1037A 

65 ft 

SW Down-gradient 




ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


29 


August 2010 



Table 4.2: TCE Monitoring Results. 


Well Identifier 

TCE (jig/L) 

Baseline 

12/12/2007 

2/19/2008 

3/19/2008 

5/19/2008 

03DP0033A 

175 

NS 

0 

212 

0 

03IJW001A 

204 

0 

0 

0 

0 

03IJW001B 

70 

0 

N/A 

N/A 

N/A 

03IJW001C 

84 

0 

N/A 

N/A 

N/A 

03MW1024A 

92 

NS 

NS 

NS 

NS 

03MW1024B 

BRL 

BRL 

1.4 

1.7 

1.7 

03MW1024C 

112 

253 

302 

0 

0 

03MW1024D 

450 

352 

0 

0 

0 

03MW1024E 

206 

0 

0 

0 

0 

03MW1024F 

184 

0 

0 

0 

157 

03MW1024G 

7.5 

0 

6.8 

6.2 

5.3 

03MW1024H 

1.6 

BRL 

0.18 

0.0 

BRL 

03MW1032A 

402 

459 

260 

447 

527 

03MW1032B 

95 

41 

48 

46 

37 

03MW1032C 

6.6 

5.9 

4.3 

4.5 

3.4 

03MW1033A 

8.2 

9 

4.4 

9.4 

13 

03MW1033B 

197 

150 

0 

0 

0 

03MW1033C 

1.3 

BRL 

BRL 

BRL 

0.0 

03MW1034A 

0.0 

0.0 

1 

0.0 

0.0 

03MW1034B 

72 

0 

0 

0 

0 

03MW1034C 

BRL 

BRL 

0.3 

BRL 

2.6 

03MW1035A 

67 

43 

80.2 

94.3 

121 

03MW1035B 

118 

151 

150 

157 

195 

03MW1035C 

4.5 

6.6 

6.9 

5.8 


03MW1036A 

148 

151 

0 

0 

0 

03MW1037A 

229 

243 

248 

219 

215 


* Purple samples were not analyzed. TCE concentration assumed to be zero. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


30 


August 2010 



































Table 4.3: Permanganate Monitoring Results. 


Well Identifier 

Permanganate (mg/L) 

11/14 

2007 

11/16 

2007 

11/20 

2007 

11/28 

2007 

Dec. 

2007* 

Jan. 

2008 

Feb. 

2008 

Mar. 

008 

May. 

2008 

03DP0033A 

NS 

NS 

NS 

NS 

NS 

258 

5.2 

0.0 

100 

03IJW001A 

NS 

NS 

NS 

NS 

3125 

6006 

8007 

7362 

4236 

03IJW001B 

NS 

NS 

NS 

NS 

55 

0.8 

0.3 

1.0 

0.5 

03IJW001C 

NS 

NS 

NS 

NS 

173 

88 

2.8 

0.0 

0.5 

03MW1024B 

0.3 

0.0 

0.3 

0.0 

0.0 

0.0 

0.3 

0.0 

0.0 

03MW1024C 

15 

0.0 

0.8 

0.8 

0.0 

0.8 

0.0 

67 

51 

03MW1024D 

0.0 

0.0 

0.0 

0.0 

0.0 

93 

103 

251 

672 

03MW1024E 

0.0 

0.0 

0.0 

0.0 

124 

3616 

4004 

2764 

904 

03MW1024F 

4.4 

258 

1472 

1731 

1033 

155 

72 

44 

0.3 

03MW1024G 

0.5 

0.0 

0.0 

2661 

982 

0.3 

0.0 

1.5 

0.0 

03MW1024H 

0.0 

0.0 

0.3 

0.0 

0.3 

0.0 

0.0 

0.0 

0.0 

03MW1032A 

3.4 

0.3 

0.3 

0.0 

0.0 

0.8 

0.5 

0.0 

0.0 

03MW1032B 

19 

0.0 

0.3 

2.8 

0.0 

0.0 

0.0 

0.0 

0.3 

03MW1032C 

7.7 

0.3 

0.0 

0.5 

0.3 

0.0 

4.4 

10.6 

0.0 

03MW1033A 

7.5 

0.0 

0.5 

0.3 

0.3 

0.0 

0.0 

0.3 

0.0 

03MW1033B 

2.6 

0.0 

0.0 

0.0 

0.0 

0.0 

1744 

2457 

51 

03MW1033C 

1.3 

0.0 

0.0 

0.3 

0.0 

0.0 

0.0 

0.0 

0.0 

03MW1034A 

15 

2.1 

1.8 

0.5 

0.5 

29 

0.8 

5.7 

0.3 

03MW1034B 

4.6 

0.0 

0.0 

0.0 

1292 

5554 

3810 

1485 

15 

03MW1034C 

2.1 

0.0 

0.0 

0.0 

0.0 

0.0 

0.0 

0.0 

0.0 

03MW1035A 

15 

1.0 

0.0 

0.5 

0.0 

0.5 

0.3 

0.0 

0.0 

03MW1035B 

3.9 

0.0 

0.3 

0.0 

0.3 

0.3 

0.0 

0.0 

0.0 

03MW1035C 

0.5 

0.3 

0.0 

0.0 

0.0 

0.3 

0.3 

0.0 

0.3 

03MW1036A 

3.6 

0.0 

0.3 

0.5 

0.3 

0.3 

21 

70 

594 

03MW1037A 

0.0 

0.0 

0.5 

0.3 

0.0 

0.0 

0.0 

0.0 

0.3 


Key: 


BRL = below reporting 
NA = not analyzed 
NS = not sampled 


imit 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


31 


August 2010 









































4.4 MODELING OF MMR PILOT TEST 


4.4.1 REACTION KINETICS 


The transport and consumption of permanganate (M) and the target groundwater contaminant (C) 
was simulated using the standard forms of the advection-dispersion equations 4-1. 


dM 


dt dx 
dt dx 


d ( D dM^ 


dx 


f 


dx j 


dx 


(vM)-r„ 


f(vc)-r c 


dx 


(4-la) 


(4-lb) 


where, 

M = Mn 04 concentration (mol L' 1 ) 

C = Contaminant concentration (mol L" 1 ) 

t = Time 

x = Distance 

D = Dispersion coefficient (L 2 T" 1 ) 

v = Pore water velocity (LT 1 ) 

Rc = linear equilibrium retardation factor of the contaminant 

r = Chemical reaction terms 


Kinetics expressions to describe the loss of permanganate (M) by NOD are based on Model 6 
from Section 2. The reaction between permanganate and the contaminant is assumed to be 
instantaneous based published reaction rates which indicate that reactions between permanganate 
and PCE or TCE are very rapid. Overall, the loss of M and C are described by the following 
three step sequence. 

(1) An instantaneous reaction between M and the contaminant (C) where 
When concentration of M > CR c Y M/c 

M - M - CR c Y m/c and C - 0 otherwise M — 0 and C = C-M /RY M/C 

(2) An instantaneous reaction between M and the instantaneous NOD (Ni) where, 
When concentration of M > Nj* p B / n 

M - M - p B * Nj / n and N t =0 otherwise M = 0 and N, = N r -M*n/p B 

(3) A second order reaction between M and slow NOD (Ns) where 

dM 1 A r / 

— p B k 2S N s M / n 
at 

^^ = -k ls N,M 

dt 2S s 

ESTCP Technical Report 
Design Tool for Planning Permanganate 
Injection Systems 32 


(4-2a) 

(4-2b) 

August 2010 



and, 

Ni = Instantaneous NOD (mol Kg' 1 ) 

N s = Slow NOD (mol Kg' 1 ) 

Pb = bulk density (Kg L' 1 ) 

Ym/c = molar ratio of M to C consumed (mol/mol) 

Ic 2 s = 2 nd order Slow NOD consumption rate (L mol' 1 d" 1 ) 

4.4.2 NUMERICAL IMPLEMENTATION 

Groundwater flow and solute transport were simulated using the MODFLOW (Harbaugh et al., 
2000) and RT3D (Clement 1997) engines within GMS (Aquaveo 2008). The chemical reactions 
between M, C, Ni and Ns were implemented through a FORTRAN 90 code compiled using 
Visual FORTRAN compiler in dynamic link libraries (rxns.dll) and formatted to fit the user- 
defined reaction package in RT3D. 

4.4.3 MODEL SETUP 

The simulation grid for the pilot test consists of 560,560 cells with 98 columns by 143 rows by 
40 layers. The cell size is non-uniform varying from 5 to 70 ft with smaller cells near the 
injection wells. Plan and profiles views of the model grid are presented in Figures 4.5 and 4.6. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


33 


August 2010 





Figure 4.6: Profile-View of MMR Pilot Test Simulation Grid. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


34 


August 2010 

































































































































































































































































































































































































































































Figure 4.7 shows the permeability distribution used in the numerical simulations. This 
permeability distribution was generated by Mr. John Glass of CH2M Hill based on multiple 
lithologic logs of the MMR site and the observed pressure response in injection and monitor 
wells during the pilot test. The higher permeability zones are represented by dark reds and lower 
permeability, silt layers by white and light pink. 



Figure 4.7: Cross-Section View of Permeability Distribution (deep red color indicates high 
permeability and white color indicates low permeability area). 


Prior to permanganate injection, TCE concentrations varied from 450 to 1.3 with an average of 
128 pg/L (standard deviation =120 pg/L). This high variability is believed to be due to spatial 
variations in hydraulic conductivity and source loading with time. The objective of this work was 
not to reproduce the contaminant loading history, so no attempt was made to calibrate the model 
to accurately match initial TCE values in individual wells. Instead, the initial TCE concentration 
for the entire model domain was set equal to the average initial TCE concentration. 

Figures 4.8 and 4.9 show plan and profile views of the injection well and monitoring well 
locations. The monitor wells are generally aligned from north-west to south-east along the 
direction of groundwater flow. The black bar in the middle of Figure 4.9 is the injection well 
and blue boxes represent well screen locations. Thin lines indicate the monitoring well cluster 
locations with screen depth indicated by the black triangles. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


35 


August 2010 











Figure 4.8: 


GW Direction 

\ 


‘03MW1032 


\ 


\ 


Injection Well 


03NIW1024-2 


03MW1035 


\ A 03MW1024-! 

\ 03MW1034 

\ 03MW1037 

\ 

\ 

03MW1036 


03MW1033 


\ 03DP0033A 

\ 

\ 

\ 


Plan-View of 15th Layer of Model Showing Injection and Monitoring Wells. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


36 


August 2010 




03MW1035 

03MW1Q34 03MVi/1032 ■ 03MW1024 {J3MW1033 03MW1037 


I 


A B 


03MW1036 


C -±A *A 


03DP0033A 


Figure 4.9: Front View of 50th Row of Model Showing Injection and Monitoring Wells 

(thick bar indicate injection well, blue box indicate well screen and triangle 
indicate monitoring well). 


Table 4.4 shows the injection rates and permanganate concentrations used in the model 
calibration. These flow rates and concentrations are based on monitoring data collected by 
CH2M Hill during the pilot test. Approximately 6% more water was injected and the injection 
period extended over five days instead of the originally planned four day period. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


37 


August 2010 


















Table 4.4: Injection Flow Rates and Concentrations Used in Model Simulations. 



Dur. 

Flow 

Rate 

MnC >4 

Cone 

Flow 

Rate 

MnCL 

Cone 

Flow 

Rate 

MnCL 

Cone 

Water 

Vol 

MnCL 

Mass 


Day 

ft 3 /d 

mol/L 

ft 3 /d 

mol/L 

ft 3 /d 

mol/L 

L 

mol 

1 st Day 

0.3743 

1,491 

0.0498 

1,592 

0.0500 

1,710 

0.0491 

50,803 

2,523 

2 nd Day 

0.4035 

1,698 

0.0000 

1,353 

0.0887 

1,763 

0.0425 

55,013 

2,406 

3 rd Day 

0.4063 

1,360 

0.0000 

1,820 

0.0605 

2,992 

0.0277 

71,013 

2,087 

4 th Day 

0.3951 

1,569 

0.0075 

1,701 

0.0059 

2,902 

0.0487 

69,042 

1,429 

5 th Day 

0.2528 

1,207 

0.0000 

1,768 

0.0000 

2,665 

0.0000 

40,376 

- 

Total 








286,247 

8,445 


* Dur. indicates duration 


Reaction parameters used in the model calibration were estimated from literature values and are 
presented in Table 4.5. 


Table 4.5: List of Common Parameters Used in Calibral 

tion Model. 

TCE Retardation Factor 

10 

- 

Molar ratio of MnCL to TCE consumed 

2 

- 

Porosity 

0.25 

- 

Bulk Density 

2 

kg/L 

Longitudinal Dispersivity 

3.28 

ft 

Horizontal Dispersivity 

0.328 

ft 

Vertical Dispersivity 

0.0328 

ft 

Molecular diffusion coefficient 

8.64E-5 

ft 2 /day 


In Section 3, NOD kinetic parameters were estimated for three soils samples from MMR based 
on a total of ten soil-mixing combinations. Overall, the kinetic parameters were similar for all 
the tests indicating a low total NOD, less than 11% instantaneous NOD, and moderate slow 
reaction rate. However, the higher permeability soil B did have a significantly higher total NOD 
(0.0054 mmol/g) than soils A and C (0.0026 to 0.0039 mmol/g). It is not clear from the batch 
test results whether the difference in total NOD between soils B and A/C is significant. 

A series of simulations were conducted to examine the sensitivity of the model simulation results 
to the NOD kinetic parameters and to determine if one parameter set provided a better fit to the 
monitoring data. Four different scenarios were evaluated. Scenarios 1, 2, and 3 use single 
values of the model parameters for the entire model domain. Scenario 1 uses parameter values 
from Soils A and C; scenario 2 uses parameter values for Soil B, and scenario 3 uses the average 
of values for Soils A, B, and C. Scenarios 4 use different values for high and low permeability 
zones in the model. High permeability zones are represented by Soil B and low permeability 
zones by Soils A and C. In the current implementation of the model, the slow reaction rate must 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


38 


August 2010 



be spatially uniform. Scenario 4 uses the averaged slow reaction rate for Soils A, B, and C. 
Table 4.6 presents the values of each parameter used in the simulations. 


Table 4.6: Details of 4 Simulation Scenarios 


Scenario 

Soil 

Type 

Fraction 

Instantaneous 

Instantaneous 

NOD 

Slow 

NOD 

Slow Reaction 
Rate 




mmol/g 

mmol/g 

L/mmol-d 

1 

A,C 

0.09 

0.0003 

0.0028 

0.0051 

2 

B 

0.06 

0.0003 

0.0051 

0.0163 

3 

A,B,C 

0.08 

0.0003 

0.0035 

0.0089 

4 

B 

0.06 

0.0003 

0.0051 

0.0107 

A,C 

0.09 

0.0003 

0.0028 


4.5 MODEL CALIBRATION 

The pilot test was simulating using four different sets of NOD kinetic parameter (see Table 4.6). 
The goodness of fit was evaluated by comparing the simulated and observed permanganate and 
TCE distributions at various time points. Two different error statistics were used: (1) root mean 
squared error; and (2) a qualitative statistic indicating the presence or absence of a compound at 
each monitoring point. 

4.5.1 SIMPLE SCORING ERROR STATISTICS (SSES) 

Permanganate reacts very rapidly with TCE. As a consequence, if permanganate reaches a zone, 
TCE is rapidly reduced to zero. Under these conditions, the presence or absence of permanganate 
in a monitoring well is often a more important indicator of treatment, than the absolute 
concentration of permanganate or TCE concentration observed in that well. 

To provide a more appropriate representation of this condition, the simple scoring error statistic 
(SSES) was developed. Using the SSES procedure, a monitor well is considered ‘contacted’ if 
the concentration of a solute (MnC >4 or TCE) is greater than 0.1% of the maximum concentration 
observed at the site. Each monitor well is then evaluated to determine if the model accurately 
matched the field observations. For example, if both the field and monitoring data indicate that 
MnC >4 is present in one well during a single sampling event, that is counted as ‘matched’. 
Conversely, if the model indicates TCE is present (at over 0.1% of max. cone.) and the field data 
indicate TCE is below detection that observation is a ‘not matched’. The fractions of matched 
observations are then reported for both TCE and MnC> 4 . 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


39 


August 2010 




4.5.2 MODEL CALIBRATION RESULTS 


Error statistics for the four simulation scenarios are presented in Table 4.7. Normalized RMSE 
is reported which is the Root Mean Square Error (RMSE) divided by the initial TCE 
concentration or the average Mn 04 concentration injected. As previously discussed, no attempt 
was made to calibrate the model to match the initial values of TCE in individual monitor wells. 
Instead, the TCE concentration for the entire model domain was set equal to the average TCE 
concentration in all wells, prior to injection. As a result, all the simulations show a high RMSE 
for TCE. To provide a basis for comparison, error statistics are also presented for a simulation 
where only water is injected with no permanganate. 


Table 4.7: Simulated and Observed Contaminant (TCE) and MnQ 4 Error Statistics. 



Contaminant (TCE) 

Mn 04 

Scenario 

Normalized RMSE 

% Matched 

Normalized RMSE 

% Matched 

Water Only 

1.00 

65% 

0.184 

77% 

1 

0.91 

82% 

0.178 

75% 

2 

0.89 

71% 

0.183 

78% 

3 

0.89 

74% 

0.180 

76% 

4 

0.89 

77% 

0.178 

77% 


In general, all the scenarios provided a reasonably good match to the field observations. 
Scenario 4 may provide a slightly better fit to the data. 

Figure 4.10 shows the simulated contaminant and permanganate distributions in a cross-section 
through the pilot test area at 6, 18, 30 and 90 days after the start of permanganate injection. 
Ground water flow is approximately right to left across the figure. The contaminant (TCE) is 
depleted (indicated by white) in the center of the pilot test area as permanganate migrates down- 
gradient. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


40 


August 2010 




TCE 


M 11 O 4 



Figure 4.10: Profile-View of Contaminant and Permanganate Distribution at 6 , 
and 90 Days of Simulation with Scenario 4 (deep red indicate high 
concentration). 


8 , 30 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


41 


August 2010 













An analysis was conducted to evaluate the sensitivity of the model simulation results to the NOD 
parameters. The pilot test was simulated using total NOD values equal to 0.1 and 10 times the 
Scenario 4 values. Error statistics for this analysis are provided in Table 4.8 for TCE and Mn 04 . 
Increasing or reducing total NOD by a factor of ten results in a poorer fit to the observed TCE 
and Mn 04 indicating the model reasonably matches the field observations. 


Table 4.8: Error Statistics Comparing Simulated and Observed Permanganate 
Measurements with Increased Total NOD. 



TCE 

Mn 04 

Scenario 

Normalize RMSE 

% Matched 

Normalized RMSE 

% Matched 

Base TNOD x 0.1 

1.04 

72% 

0.192 

69% 

Base TNOD x 1 

0.89 

77% 

0.178 

77% 

Base TNOD x 10 

0.94 

65% 

0.184 

77% 


4.6 SUMMARY AND CONCLUSIONS - MMR MODEL EVALUATION 

The numerical model RT3D was modified to simulate ISCO treatment by developing a module 
that simulates reactions between permanganate, NOD and contaminants. The reaction between 
permanganate and a single contaminant is simulated as an instantaneous reaction. Permanganate 
consumption by the natural oxidant demand is modeled assuming NOD is composed of two 
fractions: NODi which reacts instantaneously with permanganate; and NOD s which reacts with 
permanganate by a 2 nd order relationship. 

The newly developed RT3D model was used to simulate an ISCO pilot test conducted at the 
Massachusetts Military Reservation in Fall 2007 to evaluate the potential applicability of ISCO 
for treatment of the CS-10 contaminant plume. Kinetic parameters used to calibrate the model 
were estimated from prior laboratory tests. The new model provided an adequate match to the 
field data demonstrating this approach is appropriate for simulating ISCO of groundwater 
contaminants. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


42 


August 2010 




5.0 EFFECT OF INJECTION SYSTEM DESIGN ON PERFORMANCE 

5.1 INTRODUCTION 

Permanganate is often injected in a grid configuration to treat contaminant source areas. This 
grid consists of several rows of wells with multiple wells installed in each row. In some cases, 
the spacing between rows may be greater than spacing between wells within a row. This 
configuration is used when the ambient groundwater flow is used to enhance distribution of the 
permanganate. Once the target treatment zone has been defined, the designer must select a well 
spacing, mass of chemical reagent to inject, water injection volume, and injection frequency. 
Each of these parameters influences cost and remediation system performance. However, there 
is essentially no available information on the effect of these important design parameters on 
contact efficiency or the amount of permanganate released to the downgradient aquifer. 

In this project, a series of numerical model simulations were conducted to evaluate the effect of 
important design parameters on remediation system performance. Model simulations were 
performed using the numerical modeling engines MODFLOW (Harbaugh et al., 2000) 1988) and 
RT3D (Clement, 1997) within GMS (Aquaveo 2008). Permanganate consumption and 
contaminant degradation were represented using the equations presented in Section 4. 

5.2 MODEL SETUP AND BASE CASE CONDITIONS 

Figure 5.1 shows a hypothetical injection grid for treatment of a 15 m x 15 m (approximately 50 
ft x 50 ft) source area. The injection system consists of five rows of injection wells. Wells in 
each row are spaced 3.25 m on-center and rows are spaced 3 m apart (roughly 10 ft x 10 ft grid). 
Alternating rows are offset with the objective of improving reagent distribution. Unfortunately, 
it is not practical to simulate this large area with a 3-dimensional heterogeneous permeability 
distribution. To reduce the computational burden, we have chosen to simulate a subsection of 
the treatment area shown by the dashed rectangle near the center of Figure 5.1. For a uniform 
grid, this subsection can be repeated over and over again to simulate the overall treatment area. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


43 


August 2010 




Figure 5.1: Hypothetical Injection Grid Showing Model Domain Subarea. 

Figure 5.2 shows an enlarged view of the model domain subsection. Standard conditions for the 
base case simulations are summarized in Table 5.1. Overall dimensions of the simulation grid 
are 3 Sy * Sx * Z where Z is the vertical thickness of the injection zone. Contaminant and 
permanganate transport were simulated with the Third order TVD scheme (ULTIMATE) within 
the RT3D numerical model. Dispersion was simulated with the GCG Package including the full 
dispersion tensor. Chemical reaction terms were solved using the 4 th order Runge-Kutta solver. 
Grid discretization was Ax = 0.25, Ay = 0.25 m and Az = 0.05 m resulting in a 13 * 73 * 60 grid 
containing 56,160 cells. The background hydraulic gradient was established using constant head 
cells along the upgradient and downgradient boundaries of the grid. The initial PCE 
concentration was assumed to be 12,620 jig/L (0.0761 mmol L' 1 ) throughout the model domain. 
No flow boundaries are placed to simulate a recurring pattern of injection wells perpendicular to 
groundwater flow. The injection rate was 2 L/min per well (2.88 m 3 /d) for wells 1-4 and 4 L/min 
for well 5. 



Figure 5.2: Model Domain for Base Case Condition. Red shaded rectangle in center is 
target treatment zone. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


44 


August 2010 
















Table 5.1: Base Case Simulation Conditions. 


Parameter 

Value 

Units 

Well spacing perpendicular to flow (Sx) 

3.25 

m 

Well spacing parallel to flow (Sy) 

6 

m 

Vertical thickness of treatment zone (Z) 

3 

m 

Effective porosity (n) 

0.2 


Bulk Density 

2000 

Kg m' 3 

Longitudinal Dispersivity (oil) 

0.01 

m 

Transverse Dispersivity (ax) 

0.001 

m 

Vertical Dispersivity (a v ) 

0.0002 

m 

Molecular diffusion coefficient 

10" y 

m 2 s' 1 

Stochiometric coefficient (Y M /c) 

1.33 

mol MnCVmol PCE 

Slow NOD reaction rate (K s ) 

0.02 

L mmol" 1 d" 1 

Initial Contaminant Concentration 

0.0761 

mmol PCE L" 1 

Contaminant Retardation Factor (Rc) 

10 

unitless 

Horizontal correlation length (k x = X. v ) 

2.0 

m 

Vertical correlation length (X 7 ) 

0.2 

m 

Injection rate per well (Q) 

5.76 

m 3 d" 1 

Total NOD (Ni + N s ) 

0.005 

mol Kg" 1 

Ratio Instantaneous NOD to Total NOD 

0.05 



The hydraulic conductivity field was represented as a spatially correlated random field with three 
levels of heterogeneity: (a) low; (b) moderate; and (c) high (Table 5.2). Five realizations of the 
permeability distribution were simulated for each level of heterogeneity. The realizations were 
generated using the turning bands method (Tompson et al., 1989) with a horizontal correlation 
length of 2 m and a vertical correlation length of 0.2 m. Table 5.3 presents summary statistics 
for each realization. 


Parameter 

Low 

Heterogeneity 

Moderate 

Heterogeneity 

High 

Heterogeneity 

Average hydraulic conductivity 
(m/d) (Kave) 

5 

0.50 

0.05 

Variance of Ln K (a ) 

0.25 

1.0 

4.0 

Background (dh/dL) 

0.004 

0.01 

0.04 

Average Velocity (m/d) (V ave ) 

0.1 

0.025 

0.01 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


45 


August 2010 




Table 5.3: Statistical Characteristics of Ln Transformed Hydraulic Conductivity 

Distributions used in Model Simulations. 



Mean 

Variance 

Low Heterogeneity 

Realization 1 (LI) 

5.997 

0.220 

Realization 2 (L2) 

5.092 

0.189 

Realization 3 (L3) 

5.528 

0.282 

Realization 4 (L4) 

5.916 

0.224 

Realization 5 (L5) 

5.913 

0.296 

Moderate Heterogeneity 

Realization 1 (Ml) 

0.596 

1.028 

Realization 2 (M2) 

0.810 

0.879 

Realization 3 (M3) 

0.838 

0.811 

Realization 4 (M4) 

1.016 

0.984 

Realization 5 (M5) 

0.807 

0.821 

High Heterogeneity 

Realization 1 (HI) 

0.325 

3.036 

Realization 2 (H2) 

0.192 

3.211 

Realization 3 (H3) 

0.297 

3.344 

Realization 4 (H4) 

0.555 

4.374 

Realization 5 (H5) 

0.542 

4.243 


5.2.1 SCALING FACTORS 

To allow easy comparison between different simulations, the mass of reagent injected and 
volume of fluid injected are presented as dimensionless scaling factors. The volume scaling 
factor (SF V ) is the ratio of fluid (reagent plus water) injected to the pore volume of the target 
treatment zone where 

SF V = Volume of water injected / (n e Sw Sr Z) 

n e is the effective porosity and Z is the effective saturated thickness. 

The mass scaling factor (SF M ) is the ratio of reagent injected to the ultimate oxidant demand 
(UOD) of the target treatment zone where 

SF m = Mn 04 Mass injected / (UOD p B Sw Sr Z) 

UOD = NODi + NOD s + C * Rc * Y u/C / p B 

where 


NODi = Instantaneous NOD (mol Kg' 1 ) 
NOD s = Slow NOD (mol Kg' 1 ) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


46 


August 2010 




Pb = bulk density (Kg L" 1 ) 

C = Average contaminant concentration in treatment zone (mol L' 1 ) 

Rc = linear equilibrium retardation factor of the contaminant 

Ym/c = molar ratio of M to C consumed (moles/mol) 

For the base case simulations, NODi =0.004 mol Kg" 1 , NODs = 0.036 mol Kg" 1 , the initial 
contaminant (PCE) concentration = 0.0761 mmol L" 1 (12,620 pg/L), and Rc = 10. For SF M = 1.0, 
the amount of permanganate injected was 2160 mol (341 Kg) per well. However, since half of 
oxidant and fluid injected into wells on the treatment zone boundary (wells 1-4) would migrate 
outside of the model domain, the flow rate entered into the numerical model was reduced by half. 

5.2.2 TYPICAL SIMULATION RESULTS 

Figures 5.3 and 5.4 show simulated hydraulic conductivity, MnOzt, NODi, NODs, and 
contaminant distributions in both plan (Figure 5.3) and longitudinal cross-section (Figure 5.4) for 
the treatment zone subsection shown in Figure 5.2 at 120 days after the start of Mn 04 injection 
with SF V = SF m = 0.25 (injection duration = 0.25 d). In these simulations, the wells are injected 
sequentially (from 1 to 5) and the aquifer was assumed to be moderately heterogeneous 
(permeability realization #3). For SF V = SF M =0.25, sufficient water is injected to fully saturate 
25% of the pore space within the treatment zone and consume 25% of the ultimate oxidant 
demand within the treatment zone. 

In plan-view (Figure 5.3), the distribution of permanganate appears to be controlled by the 
location of the injection points, permeability distribution, and ambient groundwater flow. 
Initially, the highest permanganate concentrations develop near the injection wells. Once 
injection is complete, permanganate migrates down-gradient, preferentially migrating through 
the higher permeability zones. By 120 days, Mn 04 concentrations are low throughout the 
simulation domain due to reaction with contaminants and NOD, and down-gradient transport. 
NODi is depleted (indicated by white) in any area contacted by Mn 04 . Contaminant is also 
depleted in these areas. However, the contaminant depleted zones are slightly larger with more 
diffuse boundaries than the NODi depleted zones, due to transport of dissolved contaminant into 
areas containing Mn 04 . The areas were NOD s is depleted are much smaller than the NODi and 
contaminant depleted areas, indicating substantial amounts of NODs remain in areas that were 
temporarily contacted by Mn 04 . 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


47 


August 2010 




Figure 5.3: Horizontal Hydraulic Conductivity, MnC> 4 , NODi, NODs and Contaminant 
Distributions in Top Layer of Aquifer (see Figure 5.4) at 120 days after 
Injection for Moderately Heterogeneous Aquifer (realization #3) when Wells 
1-5 are Injected with SF V = SF M = 0.25. Deep red indicates a very high 
concentration or value, white indicates very low or zero. 

In profile-view (Figure 5.4), the effects of a heterogeneous permeability distribution on MnC >4 
transport are very apparent. In high permeability layers, MnC >4 migrates rapidly down-gradient 
resulting in complete depletion of NODi from these layers. As in plan-view, the contaminant 
depleted zones appears to be slightly larger than the NODi depleted zones, due to transport of 
contaminant from untreated areas into zones with residual Mn 04 . NODs depleted zones are 
much smaller than the NODi depleted zones indicating significant amounts of NOD s remain in 
zones that were at least temporarily contacted by Mn 04 . 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


48 


August 2010 



















Figure 5.4: Horizontal hydraulic conductivity, MnC> 4 , NODi, NODs and contaminant 
distribution in last row of aquifer (bottom row of Figure 5.3) at 120 days 
after injection for moderately heterogeneous aquifer (realization #3) when 
wells 1-5 are injected with SFV = SFM = 0.25. Deep red indicates a very high 
concentration or value, white indicates very low or zero. 

The numerical model simulations indicate that permanganate injection can be effective in 
destroying contaminants in substantial portions of the simulation domain. However, significant 
portions of the model domain may not be contacted with MnC^ and consequently, large amounts 
of contaminant remain untreated. Over time, this untreated contaminant will migrate down- 
gradient resulting in an apparent rebound in contaminant concentrations. Very high contact 
efficiencies are be required to reach target cleanup standards at many sites. In subsequent 
sections, results from a series of sensitivity analyses are presented illustrating the effect of 
different parameters on contact efficiency. This information can be used to generate improved 
designs with higher contact efficiencies. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


49 


August 2010 



















5.2.3 TREATMENT EFFICIENCY CRITERIA 


Ideally, we would like to uniformly distribute permanganate throughout the entire treatment 
zone. However, spatial variations in hydraulic gradient and hydraulic conductivity result in a 
heterogeneous permanganate distribution and less than desired treatment efficiency. Simulation 
results presented in subsequent sections show that permanganate distribution and resulting 
treatment efficiency can be enhanced by modifying the injection approach. However, modifying 
the injection approach may increase costs and/or increase the amount of unreacted permanganate 
released to the downgradient aquifer. Quantitative measures of the distribution efficiency are 
needed to evaluate the relative benefits of alternative injection approaches. In this work, three 
different measures of treatment efficiency were examined. 

The first is the Aquifer Volume Contact Efficiency (E v ) where 

Ey = volume where NODr is reduced by over 90% 
total volume of treatment zone 

The second is the Contaminant Mass Treatment Efficiency (E M ) 

E m = volume where contaminant is reduced by over 90% 
total volume of treatment zone 

The third is the unreacted (U) fraction of injected Mn0 4 

U = mass of injected MnQ 4 - AUOD 
mass of injected Mn0 4 

E v , E m and U were typically evaluated at 180 days after injection for the target treatment zone, 
defined as the region between the 1 st and 3 rd rows of injection wells (shaded area) in Figure 5.2. 
AUOD is the change in the Ultimate Oxidant Demand (UOD) from 0 to 180 days. When UOD is 
high, more Mn0 4 must be injected to treat the same volume of aquifer. In some cases, the UOD 
is so high that ISCO treatment with Mn0 4 is not practical. A MATLAB procedure was 
developed to query to model simulation results and generate statistics for E v , E M and U. 

5.3 EFFECT OF FLUID VOLUME, PERMANGANATE MASS AND TIME ON 
TREATMENT EFFICIENCY 

A series of simulations were conducted to examine the effect of injection fluid volume, Mn0 4 
mass, and time on treatment efficiency for 3-D heterogeneous conditions. Figure 5.5 shows the 
variation in Ey and Em as a function of Mn0 4 mass injected (represented by variations in SFm) 
and time. For all the simulation results presented in Figure 5.5, a total of 0.25 pore volumes of 
fluid is injected (SF V =0.25) which is equivalent to varying the Mn0 4 concentration while 
keeping the injection volume constant. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


50 


August 2010 




♦ 


-e- 

e 

a 


SFM=1.0 

SFM=0.75 

SFM=0.5 

SFM=0.25 

SFM=0.1 


Time (days) 


Figure 5.5: 



♦ 


• 0 - 

e 

A 


SFM=1.0 

SFM=0.75 

SFM=0.5 

SFM=0.25 

SFM=0.1 


Time (days) 

Variation in Aquifer Volume Contact Efficiency and Contaminant Mass 
Treatment Efficiency (Em) with Time where Fluid Injection Volume is held 
Constant (SF V =0.25) and MnC >4 Mass Varies (SF M varies from 0.1 to 1.0). 


For small amounts of MnC >4 injected (SF M = 0.1), MnC >4 is rapidly consumed and both E v and 
E m remain approximately constant over time. However, when greater amounts of MnC >4 are 
injected (SF M > 0.25), E v and E M increase gradually with time, due to down-gradient migration / 
dispersion of dissolved MnC> 4 . The increase in Ey and Em with time is greatest for the highest 
values of SF M , since larger amounts of MnC >4 would last the longest, allowing for the greatest 
drift/dispersion. The overall trends in E v and Em shown in Figure 5.5 are similar, suggesting that 
tracking both E v and E M is unnecessary. In subsequent work, only results for E v are presented 
when the impact on E v and E M is similar. However if model results indicate substantial 
differences in E v and E M , both performance measures are presented. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


51 


August 2010 






















Figure 5.6 shows the effect of varying Mn 04 mass (varying SFm) and varying fluid volume 
(varying SF V ) on E v and U at 180 days. For relatively small amounts of Mn 04 injected (constant 
SF m = 0.1 to 0.25), increasing fluid volume injected (increasing SF V ) results in a substantial 
improvement in E v . However, it also results in a substantial increase in the amount of MnCE 
flushed out of the target treatment zone (indicated by increasing fraction unreacted MnCE). For 
larger amounts of MnC >4 injected (SF M = 0.5 to 0.75), injecting large amounts of water provides 
less benefit in terms of improved contact efficiency (E v ), but also less negative impacts 
associated with migration of MnC >4 out of the target treatment zone. 




Figure 5.6: Variation in Aquifer Volume Contact Efficiency (EV) and Fraction 

Unreacted MnC >4 (U) at 180 days after Injection with Mass and Volume 
Scaling Factors. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


52 


August 2010 











For relatively small injection volumes (SF V = 0.1 to 0.25), increasing MnC >4 mass injected 
(increasing SF M ) results in a substantial increase in E v and U. This trend is due to down-gradient 
drift/dispersion of Mn 04 when larger amounts of reagent are injected. The highest volume 
contact (E v ) occurs when both SF V and SF M are increased, but also results in the release of MnCU 
outside of the target treatment zone (U). Injecting larger amounts of water (increasing SF V ) result 
in more rapid distribution of the MnC >4 before it is consumed by reaction with NODs. Injecting 
more Mn 04 (increasing SFm) allows the MnC >4 to last longer, contacting a greater portion of the 
aquifer. Overall, treatment efficiency appears to be best when SF M = SF V . This approach was 
used in subsequent analyses examining the effect of different parameters on E v . 

5.4 EFFECT OF INJECTION DESIGN PARAMETERS ON PERFORMANCE 

The simulation results presented above demonstrate that increasing injection fluid volume and 
mass of Mn 04 increases contact and treatment efficiency. Injection well spacing and reinjection 
frequency can also be adjusted to improve efficiency. 

Figure 5.7 shows the effects of varying the injection well spacing on E v and U at 180 days when 
the well spacing in a row is 3.25 m or 6.25 m, and the spacing between rows of injection wells is 
3, 6, 9, or 12 m. In these simulations, the amount of solution injected per well is varied so that 
the total volume of fluid and mass of MnC >4 injected per volume of treatment zone (SF V and 
SF m ) remained constant for different well and row spacings. For both 3.25 and 6.25 m well 
spacing, increasing the row spacing had a modest impact on volume contact efficiency (E v ) and 
fraction unreacted MnC >4 (U) when SF V =SF M was less than 0.25. However for larger values of 
SF v =SF m (>0.5), increasing the row spacing had a substantial negative impact, reducing volume 
contact efficiency and increasing the amount of unreacted MnC >4 released downgradient. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


53 


August 2010 




Figure 5.7: Effect of Well Spacing on Ey and U at 180 days for SFy = SFm in a Medium 
Heterogeneity Aquifer. 

Figure 5.8 shows the effects of performing multiple injections when the total fluid and mass of 
MnC >4 is held constant. Results are presented for E v at 180 days after the first injection. In these 
simulations, the volume of injection solution was split into multiple injection events while the 
total volume of material injected remained constant. For a single injection and SF v =SFm= 1, 5.8 
m 3 of 0.405 M MnC >4 solution (100% injection) was injected into each well on Day 0. For the 
four injections and SF v =SFm= 1, 1.45 m 3 of 0.405 M MnC >4 solution (25% injection) was injected 
into each well on Days 0, 45, 90, and 135. 

Many practioners apply MnC >4 through multiple injection events in an effort to increase contact 
efficiency. Results presented in Figure 5.8 indicate that, if the total volume of fluid injected and 
mass of reagent is held constant, multiple injection events through the same wells will not 
increase contact efficiency compared to a single large injection event, and can increase the 
amount of unreacted MnC >4 released downgradient. This occurs because several small injections 
through the same wells repeatedly treat the same area around the wells, with little increase in the 
fraction of aquifer contacted. However if the injection wells or points are relocated after each 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


54 


August 2010 













injection, multiple injections will improve contact since this would be similar to using a smaller 
well spacing. 



00 


50 


0 0.25 0.5 0.75 

Scaling Factor (SF\/ = SFm) 


0 0.25 0.5 0.75 

Scaling Factor (SFv = SFm) 


Figure 5.8: Effect of Reinjection on E v and U at 180 days for SF V = SF M in a Medium 

Heterogeneity Aquifer. Well spacing = 3 m. 

Users should be aware that at many sites, it may not be practical to inject a large amount of water 
in a single injection due to pressure buildup in the aquifer and multiple injections are required. 
Results shown in Figure 5.8 indicate that several small injections will significantly improve 
performance compared to a single small injection. For example, four injections of 0.25 pore 
volumes each is estimated to have a volume contact efficiency (E v ) of 71% compared to one 
0.25 PV injection which would have an E v of 39%. 

5.5 EFFECT OF SITE CHARACTERISTICS ON PERFORMANCE 

A series of sensitivity analyses were conducted to examine the effect of model input parameters 
on contact efficiency (E v ). The parameters expected to have the greatest impact on E v were 
thought to be: (1) the Slow NOD reaction rate (kis)', (2) the Total NOD (Ni + Ns); (3) the ratio of 
NODi to Total NOD; (4) the initial contaminant concentration; (5) the contaminant retardation 
factor (R); and (6) the level of heterogeneity. Table 5.4 shows the parameter values used the 
sensitivity analyses. In all simulations, there was a single Mn 04 injection with SFy = SFm. 

Figure 5.9 shows the effect of varying the slow NOD reaction rate (&?s), total NOD, and fraction 
NODi on E v . Lower values of /q.s and total NOD result in an increase in both E v and U due to 
slow consumption of Mn 04 . In contrast, the fraction of NODi had negligible impact on Ev and U. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


55 


August 2010 




Table 5.4: Input Parameters used in Sensitivity Analyses Simulations. 



Ratio 
to Base 

Initial C 

Total 

NOD 

NODi 

k2S 

Fraction 

NODi 

R 

MnCL 
Injectio 
n Cone. 

K a ye 

Back¬ 

ground 

Gradien 

t 

(dh/dL) 

V a ve 



mmol/L 

mol/kg 

mol/kg 

L/mol-day 

% 

- 

mol/L 

m/day 

- 

m/day 

Slow reaction rate 

0.1 

0.0761 

0.005 

0.00025 

0.002 

5 

10 

0.055 

0.5 

0.01 

0.025 

1 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.5 

0.01 

0.025 

2.5 

0.0761 

0.005 

0.00025 

0.05 

5 

10 

0.055 

0.5 

0.01 

0.025 

Total NOD 

0.2 

0.0761 

0.001 

0.00005 

0.02 

5 

10 

0.015 

0.5 

0.01 

0.025 

1 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.5 

0.01 

0.025 

4 

0.0761 

0.02 

0.001 

0.02 

5 

10 

0.205 

0.5 

0.01 

0.025 

Fraction 

Instantaneous NOD 

0.2 

0.0761 

0.005 

0.00005 

0.02 

1 

10 

0.055 

0.5 

0.01 

0.025 

1 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.5 

0.01 

0.025 

2 

0.0761 

0.005 

0.0005 

0.02 

10 

10 

0.055 

0.5 

0.01 

0.025 

Initial Contaminant 
Concentration 

0.1 

0.00761 

0.005 

0.00025 

0.02 

5 

10 

0.051 

0.5 

0.01 

0.025 

1 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.5 

0.01 

0.025 

10 

0.761 

0.005 

0.00025 

0.02 

5 

10 

0.101 

0.5 

0.01 

0.025 

Retardation Factor 

0.1 

0.0761 

0.005 

0.00025 

0.02 

5 

1 

0.051 

0.5 

0.01 

0.025 

1 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.5 

0.01 

0.025 

10 

0.0761 

0.005 

0.00025 

0.02 

5 

100 

0.101 

0.5 

0.01 

0.025 

Heterogeneity 

Level 

LOW 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

5 

0.004 

0.1 

MID 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.5 

0.01 

0.025 

HIGH 

0.0761 

0.005 

0.00025 

0.02 

5 

10 

0.055 

0.05 

0.04 

0.01 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


56 

August 2010 









Figure 5.9: Effect of NOD Kinetic Parameters on E v and U at 180 days for SFy=SF M : 

(a) Slow NOD Reaction Rate (k 2 s); (b) Total NOD; and (c) Fraction NOD L 

Figure 5.10 shows the effect of varying the initial contaminant concentration and contaminant 
retardation factor on E v? E M and U. Both parameters had a minor impact on E v . However, 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


57 


August 2010 











contaminant retardation factor had a large impact on EM, presumably due to increased contact 
between the relatively immobile contaminant and MnCE. High values of initial contaminant 
concentration and retardation factor increased fraction of MnC >4 released downgradient due to the 
larger mass of MnCE injected. 





Figure 5.10: Effect of Initial Contaminant Concentration (a) and Contaminant 
Retardation Factor (b) on E v , Em, and U at 180 days for SFv=SFm. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


58 


August 2010 








5.5.1 EFFECT OF AQUIFER HETEROGENEITY ON E M 


Figure 5.11 shows the effect of varying the level of aquifer heterogeneity on E v , E M and U. The 
error bars represent the standard deviation of these values obtained for the five different 
realizations simulated for each level of heterogeneity (see Table 5.2 and 5.3). For a given 
volume of Mn 04 solution injected, aquifer volume contact efficiency (E v ) is highest and U is 
lowest for the low heterogeneity permeability distribution. E v decreases and U increases as the 
level of heterogeneity increases. This is due to flow bypassing around lower permeability zones, 
causing these zones to remain uncontacted. 

Aquifer heterogeneity appears to have a complex interaction with contaminant mass removal. 
E m is highest for the moderate and high heterogeneity simulations. For volume and mass scaling 
factors (SF V and SF M ) less than 0.75, the medium and high heterogeneity distributions result in a 
somewhat higher treatment efficiency than the low heterogeneity distribution. However, for 
SF v =SFm=1, Em is lower for the high heterogeneity simulations than the moderate and low 
heterogeneity simulations. 

The presence of higher permeability channels in the moderate and high heterogeneity simulations 
may allow more rapid transport of Mn 04 away from the injection well when only a small amount 
of fluid is injected. Once Mn 04 is distributed throughout the treatment zone, contaminants can 
diffuse out of lower permeability layers and come in contact with Mn 04 , somewhat increasing 
E m . However for large values of SF V and SF M , permanganate cannot penetrate the lower 
permeability zones, reducing the maximum level of treatment that can be achieved. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


59 


August 2010 



Figure 5.11 



Scaling Factor (SFv = SFm) 


Effect of Low, Medium and High Aquifer Heterogeneity on E v , E M and U at 
180 days (SF V = SF M ) (error bars show the standard deviation observed in five 
permeability realizations for each level of heterogeneity) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


60 


August 2010 





















5.6 SUMMARY AND CONCLUSIONS - EFFECT OF INJECTION SYSTEM 

DESIGN VARIABLES AND SITE CHARACTERISTICS ON REMEDIATION 
SYSTEM PERFORMANCE 

The ISCO numerical model developed in Section 3 was applied to a hypothetical heterogeneous 
aquifer to evaluate the effect of different design variables and aquifer parameters on treatment 
efficiency. Model simulation results indicate that the two parameters with the greatest impact on 
aquifer contact efficiency and pollutant treatment efficiency are: (1) the mass of permanganate 
injected; and (2) the volume of water injected. To allow easy comparison between different 
simulations, the mass of permanganate injected and volume of fluid injected were represented as 
dimensionless scaling factors. The volume scaling factor (SF V ) is the ratio of fluid volume 
injected to the pore volume of the target treatment zone. The mass scaling factor (SF M ) is the 
ratio of reagent injected to the ultimate oxidant demand (UOD) of the target treatment zone. 
When small amounts of permanganate are injected, the reagent is rapidly consumed and pollutant 
removal efficiency does not increase with time after the first 30 days. However, when larger 
amounts of permanganate are injected, the reagent can persist for several months resulting in a 
gradual increase in treatment efficiency with time. For constant permanganate mass, increasing 
fluid volume injected initially results in an improvement in treatment efficiency. However, 
further increases in fluid volume injected result in little additional benefit. Conversely, when 
fluid volume is held constant and permanganate mass is increased, treatment efficiency steadily 
increases, due to down-gradient drift/dispersion of permanganate when larger amounts of reagent 
are injected. However, increasing mass of MnC >4 injected, also incrases the amount of MnC >4 that 
migrates out of the target treatment zone (indicated by increasing fraction unreacted MnOzO. 

Common approaches for improving remediation system performance include: (a) reducing the 
injection well spacing; and (b) performing multiple MnC >4 injections in the same well. 

Numerical model simulation results indicate that well spacing within a row, and row spacing 
have a modest impact on volume contact efficiency (E v ) and fraction unreacted MnC >4 (U) for 
small amounts of MnC >4 and fluid injected (SF V and SF M < 0.25). However for larger amounts of 
reagent and fluid injected, increasing row spacing had a substantial negative impact, reducing 
volume contact efficiency and increasing the amount of unreacted MnC >4 released downgradient. 

Numerical model simulation results indicate that, if the total volume of fluid injected and mass of 
reagent is held constant, multiple injection events will not increase contact efficiency compared 
to a single large injection event, and can increase the amount of unreacted MnC >4 released 
downgradient. This occurs because several small injections through the same wells repeatedly 
treats the same area around the wells, with little increase in the fraction of aquifer contacted. 
This occurs because several small injections result in more complete removal of NOD s near the 
injection well, compared to a single large injection where the MnC >4 is rapidly transported away 
from the injection well. However at many sites, it may not be practical to inject a large amount 
of water in a single injection due to pressure buildup in the aquifer and multiple injections are 
required. In these cases, multiple small injections can significantly improve performance 
compared to a single small injection. 

A series of sensitivity analyses were conducted to examine the effect of site characteristics on 
remediation system performance. The initial contaminant concentration and contaminant 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


61 


August 2010 



retardation factor had only a minor impact on remediation system performance, but did have a 
large impact on the fraction of Mn 04 released downgradient. Lower values of kjs and total NOD 
result in an increase in both Ey and U due to slow consumption of Mn 04 . In contrast, the 
fraction of NODi had negligible impact on E v and U. 

Aquifer heterogeneity has a complex relationship with remediation system performance. 
Increasing levels of heterogeneity have a negative effect on aquifer volume contact efficiency 
(E v ) and fraction of Mn 04 released downgradient (U). However, moderate and high levels of 
heterogeneity appear to increase contaminant mass treatment efficiency (E M ) when only small 
amounts of MnCL solution are injected. However, when large amounts of MnCL solution are 
injected, E M is lower for high heterogeneity simulations. The presence of higher permeability 
channels in the moderate and high heterogeneity simulations may allow more rapid transport of 
MnCL away from the injection well when only a small amount of fluid is injected, improving 
contaminant treatment. However, permanganate cannot penetrate the lower permeability zones 
in highly heterogeneous formations, reducing the maximum level of treatment that can be 
achieved. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


62 


August 2010 



6.0 SPREADSHEET BASED MODELING OF PERMANGANATE 
DISTRIBUTION 

6.1 INTRODUCTION 

Permanganate injection can be very effective for ISCO of a variety of ground water 
contaminants. However to be effective, the permanganate must be brought into close contact 
with the target contaminant. Simulation results presented in Section 4 indicate that aquifer 
volume contact efficiency will be controlled by a variety of factors including the mass of 
permanganate injected, water volume injected, aquifer heterogeneity and kinetics of 
permanganate consumption by NOD. Detailed injection system designs can be evaluated using 
the ISCO module developed and tested in Section 3. However, use of this advanced model is not 
practical in many cases due to time, budget and data constraints. Simple, easy to use tools are 
needed to assist designers in developing lower cost, more effective remediation systems. 

In this section, we describe the development and implementation of an Excel-based spreadsheet 
tool to plan CDISCO. Permanganate transport and consumption in a homogeneous aquifer are 
simulated using the kinetic approach developed and applied in Sections 3 and4. The effective 
ROI is then determined based on a user specified contact time and critical permanganate 
concentration. The ROI is then used to determine required well spacing, injection parameters 
and generate preliminary cost estimates. Injection parameters can be quickly changed allowing 
designers to easily evaluate multiple alternatives, allowing designers to quickly identify lower 
cost, more effective designs. 

This model is implemented as part of the CDISCO design tool developed with support from the 
Environmental Technology Certification Program (ESTCP) under project ER-0623 and ER- 
0625. 

6.2 SIMULATING OXIDANT DISTRIBUTION USING A SERIES OF CSTRS 
6.2.1 MODEL DEVELOPMENT 

The transport and consumption of permanganate are simulated within the CDISCO model using 
the same differential equations developed and tested in Section 3. Based on user input data, the 
model establishes a series of continuously stirred tank reactors (CSTR) to simulate advective- 
dispersive transport of permanganate away from a central injection well. Formally, this is a 
special case of the fully upwinded finite difference solution corrected for numerical dispersion 
presented by Van Genuchten and Wierenga (1974). Numerical dispersion associated with 
evaluation of the time derivative is negligible (Van Genuchten and Wierenga, 1974) so 
longitudinal dispersion is simulated by setting the length of each reactor to two times the 
longitudinal dispersivity (oil). The volume of each CSTR reactor increases outward from the 
injection well, simulating the decline in velocity during radial flow. The volume of the first 
reactor (Vi) is B e n n (a L ) 2 where B e is the effective saturate thickness and n is porosity. The 
volume of each additional reactor V N is B e n n [(aL +2(N-1) aL)) - (aL +2(N-2) aL)) ] where N is 
the reactor number. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


63 


August 2010 



The CSTR model is implemented within a MS Excel spreadsheet. The user first enters 
information on aquifer characteristics (porosity, hydraulic conductivity, injection interval, total 
NOD, contaminant concentrations, etc.), injection conditions (permanganate injection 
concentration, flow rate and duration), and target conditions (minimum oxidant concentration 
and duration to calculate ROI). The model automatically converts the input parameters to 
appropriate units. Model parameters and units for data entry are: 

Initial contaminant concentration (mg/L) 

Oxidant concentration injected (mg/L) 

Total NOD (g/Kg) 

Fraction instantaneous NOD (dimensionless) 

Molar ratio of contaminant to oxidant consumed (mmol/mmol) 

Contaminant retardation factor 
Aquifer bulk density (Kg/L) 

Aquifer porosity (dimensionless) 

2 nd order slow NOD consumption rate (L/mmol-d) 

Longitudinal dispersivity (ft) 

Effective saturated thickness (ft) 

Permanganate transport and consumption are simulated using the stepwise calculations shown 
below. The actual computations are performed in a Visual Basic Macro embedded in the 
CDISCO spreadsheet. The time step for each computation is automatically determined within 
Excel to minimize computational error. 

Step 1 Advective-Dispersive Transport of Oxidant and Contaminant 

The change in concentration of oxidant (M) and contaminant (C) is calculated for each reactor by 
the following equations. 

™=^-M)Q>v 

^(C,-C)QIVR 


where Q is the injection rate. 

Step 2 Instantaneous Reaction of Oxidant with Contaminant 

When excess oxidant is present, the contaminant concentration is set to zero and the oxidant 
concentration is reduced by an amount equal to: 


C * Y M /c * R 


When excess contaminant is present, the oxidant concentration is set to zero and the contaminant 
concentration is reduced by an amount equal to: 

M/(Ym /C * R) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


64 


August 2010 



Mathematically, this is expressed as: 


If M‘ > CRY m/c , then [ M m = M' - C‘RY M/c and C M = 0] 
else [C ,+l = C - M' / RY m/c and M i+X = 0] 

Step 3 Instantaneous Reaction of Permanganate with Soil and Groundwater NOD 

When excess oxidant is present, the instantaneous NOD concentration is set to zero and the 
oxidant concentration is reduced by an amount equal to: 

Ni Ps / n 

When excess Ni is present, the oxidant concentration is set to zero and the NI is reduced by an 
amount equal to: 

Nj n / p B 

Mathematically, this is expressed as: 

If M‘ > Njp B / n, then [m m = M‘ - N/p B / n and N/ +l = 0] 
else [7V/ +1 = Nj -M i nt p B and M m = 0] 

Step 4 Slow Consumption of Oxidant with Soil and Groundwater (NOD s ) 

The concentration of oxidant is further reduced by consumption from slower, long-term reaction 
with soil and groundwater. The rate of oxidant loss with time (dM/dt) is proportional to the 
oxidant concentration times the slow NOD (Ns) concentration. The amount of NOD undergoing 
this slower reaction is also reduced by this same reaction. Bulk density and water filled porosity 
are included in the equations to convert from aqueous concentrations to soil concentrations. 
Mathematically, this is expressed as: 

-= -k s N s Mp B /n 

dt 

dN s , , r ,, 

—— = -k s N s M 
dt 

6.2.2 MODEL VALIDATION 

The CDISCO spreadsheet model was validated by comparing results the CDISCO simulation 
results with: (1) analytical and RT3D simulations of non-reactive radial flow in a 2-D 
homogeneous aquifer; and (2) RT3D simulations of reactive transport using the ISCO module 
developed in Section 3. Input parameters for the non-reactive and reactive simulations are 
shown in Table 6.1. For the non-reactive RT3D simulations, NODi, NODs and initial 
contaminant concentration were set to zero so Mn 04 would transport as a conservative tracer. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


65 


August 2010 



The analytical solution used for comparison with CDISCO was developed by Gelhar and Collins 
(1971). 

Table 6.1: Base Model Parameters for Comparison of CDISCO, Analytical and RT3D 


Simulations. 


Parameter 

CDISCO 

Analytical 

RT3D- 

Reactive 

Vertical thickness of treatment zone (m) 

1.0 

1.0 

1.0 

Porosity (n) 

0.2 

1.0 

0.2 

Bulk Density (Kg/m 3 ) 

2000 

na 

2000 

Longitudinal Dispersivity (m) 

0.1 

0.1 

0.1 

Total NOD (mmol/Kg) 

50 

na 

50 

Fraction Instantaneous NOD 

0.1 

na 

0.1 

2 nd Order Slow NOD Consomption Rate (L / mmol-d) 

0.02 

na 

0.02 

Mn 04 Molecular Weight (g/mol) 

118.94 

na 

118.94 

Mn 04 Injection Concentration (mg/L) 

5000 

5000 

5000 

Injection rate per well (m /d) 

20 

120 

20 

Injection duration (d) 

1.0 

1.0 

1.0 


na - not applicable 


Figure 6.1 shows simulated solute concentrations with radial distance from the injection well for 
longitudinal dispersivity (ul) equal to 0.1 and 1.0 m at one day after injection. CDISCO 
provides a very close match with the analytical and RT3D simulation for both values of a L . 
However, there are far fewer data points for comparison with oil equal to 1.0 m because CDISCO 
automatically sets the reactor cell size equal to 2 (Xl. RMSEs were determined by comparing 
concentrations computed by CDISCO with the analytical and RT3D simulations for ol equal to 
0.1, 0.3, 0.5 and 1.0 m (Table 6.2). Overall, CDISCO provides an excellent match to the RT3D 
simulation results. The match with the analytical solution is not as good, especially for large 
values of a L . This discrepancy may be due to a limitation in the analytical solution when the 
value of ^ approaches the radial transport distance. For example, the analytical solution 
computes a concentration of 4635 mg/L at the injection well when 5000 mg/L is injected. 



ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


66 


August 2010 





















Radial Distance [m] 

Figure 6.1: Comparison of CDISCO, RT3D and ID Analytical Solutions of Non-Reactive 
Solute Transport Away from a Single Injection Well at 1 day after Injection. 

Table 6.2: Comparison of CDISCO, Analytical and RT3D Non-Reactive Simulations. 



RM 

[SE 

Longitudinal Dispersivity (m) 

0.1 

0.3 

0.5 

1.0 

Analytical - CDISCO 

0.94% 

1.59% 

3.41% 

4.59% 

RT3D - CDISCO 

0.45% 

0.76% 

1.53% 

2 . 02 % 


Figure 6.2 shows simulated M 11 O 4 , NODi, NODs, and contaminant concentrations versus radial 
distance from the injection well at 10 days after injection. CDISCO provides an excellent match 
with the RT3D simulation. This should not be surprising since the same kinetic expressions are 
incorporated into both models. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


67 


August 2010 














NODj 



NOD, 


DU 

O 

a 

a, 

C/5 

Q 

O 

Z 



Figure 6.2: Comparison of CDISCO and RT3D Simulations of MnC> 4 , NODi, NODs and 
Contaminant Concentration at 10 days after Injection for cil =0.1 m. 

6.3 COMPARISON OF CDISCO WITH 3D HETEROGENEOUS SIMULATIONS 

Simulation results presented in Section 4 showed that aquifer heterogeneity can have a 
significant impact on remediation system performance. CDISCO is a 1-D homogeneous radial 
flow model. Given this model structure, there is no way to simulate the spatial variability 
present in a 3D heterogeneous aquifer. However it might be feasible to calibrate CDISCO to 
simulate the average behavior of a 3D heterogeneous aquifer. 

To evaluate the ability of CDISCO to simulate condition in a 3-D heterogeneous aquifer, the 
RT3D model previously applied in Section 4 was used to simulate radial flow from a single 
injection well in a 20 m x 20 m x 1 m thick, heterogeneous aquifer. Figure 6.3 shows simulated 
concentrations in each cell generated in the 3-D heterogeneous simulation vs radial distance from 
the injection well. The cloud of concentration values between 5 and 10 m illustrates the 
tremendous variability in simulated solute concentrations generated by the 3D RT3D model. At 
5 m from the injection well, the spatially averaged concentration is 335 mg/L. However, 
concentrations in individual wells vary from zero to 600 mg/L. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


68 


August 2010 
































Radial Dist inct (m) 

Figure 6.3: Non-Reactive Solute Concentrations versus Radial Distance from Injection 
Well Generated in 3-D Heterogeneous RT3D Simulation. Symbols are 
Concentrations in Individual Model Cells. Line is Spatially Averaged 
Concentration (3 m radial increments). 

CDISCO was then calibrated to match the spatially averaged values from RT3D by adjusting 
longitudinal dispersivity (a L ) to a best fit value of 1.5 m. Figure 6.4 shows a comparison of the 
spatially averaged concentrations generated by RT3D with the non-reactive solute breakthrough 
curve generated by CDISCO for a L = 1.5 m. Overall, the calibrated CDISCO model provides an 
excellent fit the spatially averaged RT3D results. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


69 


August 2010 









Figure 6.4: Comparison of CDISCO Simulation (a L = 1.5 m) and Spatially Averaged 
Concentrations from 3D RT3D Simulation. 

CDISCO and RT3D were both used to simulate Mn 04 injection into a contaminated aquifer. 
Figure 6.5 shows a comparison of the spatial distribution of Mn 04 , NODi, NODs and 
contaminant generated by CDISCO and RT3D. CDISCO was run with o,l = 1.5 m. RT3D was 
run with a L = 0.01 m and a spatially heterogeneous permeability distribution. All other 
parameters were the same for the CDISCO and RT3D simulations (see Table 6.1). The 
individual points shown in Figure 6.5 are vertically averaged concentrations. 

Overall, CDISCO provided a reasonably good match to the vertically averaged RT3D simulation 
results. CDISCO accurately predicted the general shape and mid-point of the Mn 04 and NOD s 
distributions. However, CDISCO did a poor job of simulating the contaminant distribution. The 
high effective dispersivity (oil = 1.5 m) used in CDISCO results in very effective mixing of 
Mn 04 and the contaminant resulting in rapid contaminant degradation. However, in the high 
resolution RT3D simulations, mixing between the Mn 04 and contaminant is much more limited 
resulted in much more limited contaminant degradation. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


70 


August 2010 









Radial Distance (in) 


0.04 

3 

0.03 

i 0.02 

l-H 

0.01 

0.00 

0 5 10 15 20 



Radial Distance (m) 

NOD s 



— CDISCO - 
Homogeneous 

RT3D - 

Heterogeneous 


Radial Distance (in) 


Radial Distance (in) 


Figure 6.5: Comparison of Model Results at 10 days after Injection for 1-D 

Homogeneous CDISCO simulation and 3-D spatially Heterogeneous RT3D 
Simulation for: (a) Mn 04 ; (b) NODi; (c) NODs and Contaminant. 

6.4 CDISCO MODEL STRUCTURE 

The CSTR model is implemented within a MS Excel spreadsheet. On the first worksheet, users 
enter information on aquifer characteristics (porosity, hydraulic conductivity, injection interval, 
NOD, contaminant concentrations, etc.), injection conditions (permanganate injection 
concentration, flow rate and duration), and target conditions (minimum oxidant concentration 
and duration to calculate ROI). The CSTR model described above then computes the spatial 
distribution of Mn 04 , NODi, NODs and contaminant concentration as a function of radial 
distance at various times. Figure 6.6 shows typical output from the permanganate transport 
model simulations. The graph shows the variation in permanganate concentration versus 
distance at several different times (15, 30, 45 and 60 days for this simulation). The table at the 
bottom shows input parameters for a series of prior simulations. The effective Radius of 
Influence (ROI) is computed for a user specified time and minimum oxidant concentration. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


71 


August 2010 

































Oxidant Concentration vs. Radial Distance 



Radial Distance (ft) 


Selected 

Run 

Number 

Injection 

Duration 

Aquifer 

Thickness 

Thickness of 

Mobile Zone 

NOD Fraction 

Instantaneous 

NOD (g/kg) 

Slow NOD 

Rate 

Injection 

Oxidant 

Cone 

(mg/L) 

Injection 

Rate 

gal/Day 

ROI 

Minimum 

Oxidant 

Cone to 

Calc ROI 

Target 
Number of 
Days to Calc 
ROI 

* 

1 

3 

20 

10 

0.1 

5 

0.001 

10,000 

3,000 

10.41 

50 

30 

* 

2 

5 

20 

10 

0.1 

5 

0.001 

10,000 

3,000 

14.46 

50 

30 


3 

10 

20 

10 

0.1 

5 

0.001 

10,000 

3,000 

21.25 

50 

30 


4 

10 

20 

10 

0.1 

5 

0.001 

5,000 

3,000 

15.56 

50 

30 

✓ 

5 

10 

20 

10 

0.1 

5 

0.001 

20,000 

3,000 

27.23 

50 

30 

* 

6 

5 

20 

10 

0.1 

5 

0.001 

20,000 

3,000 

19.64 

50 

30 


Figure 6.6: Typical Output from Permanganate Transport Model. 


CDISCO also includes a simplified procedure to generate preliminary cost estimates based on 
the user specified treatment area dimensions, injection ROI overlap (%), number of injection 
events planned, fixed cost, and unit costs for injection point installation, chemical reagents, and 
labor for injection. Two injection approaches are feasible - injection through direct push rods or 
through wells. Cost factors are included for mobilization, labor, materials, equipment rental, 
travel, and subcontractor costs. 


Figure 6.7 shows typical output from the cost estimating procedure comparing preliminary 
estimates for several different design alternatives. Aquifer parameters, treatment area dimensions 
(100 ft x 100 ft), ROI overlap (25%), time to calculate ROI (30 days), minimum oxidant 
concentration (50mg/L), and number of injection events (5) are constant for all alternatives. 
Alternative 1 has a relatively high cost because the short injection duration (3 days) required a 
large number of injection points. Alternative 5 has a lower cost because the longer injection 
duration (10 days) and higher oxidant concentration (20,000 mg/L KMnOzj) reduced number of 
injection points required. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


72 


August 2010 






















Figure 6.7: Typical Output from Injection Scenario Cost Comparison. 

6.5 EFFECT OF OVERLAP FACTOR ON CONTACT EFFICIENCY 

All of the input parameters required by CDISCO to simulate Mn 04 transport are based on 
physical parameters that can be independently measured in laboratory or field tests. However, 
converting the simulated Mn 04 distribution into a remediation system design does require some 
judgment. 


CDISCO calculates the effective ROI of a single injection based on a minimum oxidant 
concentration (MinOx) and contact time (CT) specified by the user. The assumption is that if a 
certain minimum amount of oxidant is distributed throughout the ROI for minimum contact time, 
all of the aquifer within the ROI will be effectively treated. The user also enters an ROI overlap 
factor (OF) to account for uncertainty/variability in aquifer characteristics. OF is defined as 
2*ROI/well spacing 

Currently, there is essentially no guidance on what Mn 04 concentration, contact time, and ROI 
overlap factor is required for good treatment. In theory, even a small excess of Mn 04 should 
rapidly degrade any PCE or TCE present. However, this assumes perfect mixing of the Mn 04 
and the contaminant. As the prior simulations illustrate, mixing is often poor and there may be a 
significant benefit to providing some excess Mn 04 or having overlapping injection zones. 

A statistical analysis was performed to determine if there was a significant correlation between 
the user defined parameters (MinOx, CT, and OF) and remediation system performance. For 
each of the 3-D RT3D simulations conducted in Section 5, CDISCO was used to calculate an 
ROI based on MinOx = 100, 20 and 10 mg/L; and CT = 180, 120, 90 and 60 days. Overall, 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


73 


August 2010 






















































































computed ROIs were insensitive to the value of MinOx and MinDur used in the analysis. All of 
cases showed that correlation over 0.95. 

Figure 6.8 illustrates the relationship between aquifer volume contact efficiency (E v ), 
contaminant mass treatment efficiency (Em) and the OF using MinOx =10 mg/L and CT =180 
days. Values of OF between 1.0 and 1.5 generally resulted in high E v and Em, indicating values 
in this range will result in good remediation system performance. 



Figure 6.8: Effect of Overlap Factor (OF) on Aquifer Volume Contact Efficiency (Ey) 


and Contaminant Mass Treatment Efficiency (E M ). 

6.6 SUMMARY 

An Excel spreadsheet based model CDISCO was developed to simulate the injection and radial 
distribution of Mn 04 in aquifers undergoing ISCO. Comparisons with analytical and numerical 
models or non-reactive and reactive transport demonstrated that CDISCO accurate simulates 
Mn 04 transport and consumption. Comparisons with 3-D heterogeneous RT3D simulations 
indicates CDISCO provides reasonably good estimates of the average Mn 04 transport distance in 
heterogeneous aquifers. However, CDISCO will under estimate the maximum Mn 04 transport 
distance in higher permeability layers. 

CDISCO can be used to design Mn 04 injection systems. The primary model inputs are the 
aquifer characteristics (porosity, hydraulic conductivity, injection interval, NOD, contaminant 
concentrations, etc.), injection conditions (permanganate injection concentration, flow rate and 
duration), and unit costs for reagent, drilling and labor. However, CDISCO also requires the 
user to specify: (a) the MinOx for effective treatment; (b) minimum CT for effective treatment; 
and ROI OF. Comparison of CDISCO results with the 3-D heterogeneous simulation results 
generated in Section 4 indicate that: (a) aquifer volume contact efficiency (Ey) and contaminant 
mass treatment efficiency (E M ) are not strongly influenced by values selected for MinOx or CT; 
and (b) acceptable results are generated with MinOx =10 mg/L and CT =120 and 180 days. 
Comparison with the 3-D simulations also showed that: (a) E v and Em were sensitive to OF; and 
(b) values of OF between 1.0 and 1.5 generally resulted in good remediation system 
performance. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


74 


August 2010 





7.0 REFERENCES 


AFCEE, (2007a) The Installation Restoration Program at the Massachusetts Military Reservation 
(MMR), Air Force Center for Engineering and the Environment (http://m 1 nr. 0 rg/t . 

AFCEE, (2007b) Groundwater Plume Maps and Information Booklet, Air Force Center for 
Engineering and the Environment. 

ATSDR, (2004) Medical Management Guidelines for Tetrachloroethylene, Agency for Toxic 
Substances and Disease Registry, Division of Toxicology and Environmental Medicine, U.S. 

ATSDR, (2007) CERCLA Priority List of Hazardous Substances that will be The Subject of 
Toxicological Profiles and Support Document, Agency for Toxic Substances and Disease 
Registry, Division of Toxicology and Environmental Medicine, U.S. 

ASTM D7262, (2007) Standard Test Method for Estimating the Permanganate Natural Oxidant 
Demand of Soil and Aquifer Solids, In; ASTM International: West Conshohocken, Pennsylvania 


CH2M Hill, (2007) Project Note for CS-10 ISCO Pilot Test Work Plan. 

Chambers, J.D., A.L. Leavitt, C.L. Walti, C.G. Schreier, J.T. Melby, and L. Goldstein, (2000) 
Treatability Study - Fate of Chromium during Oxidation of Chlorinated Solvents, In: 

Proceedings of International Conference on Remediation of Chlorinated and Recalcitrant 
Compounds, Columbus, Ohio, pp. 57-66. 

Clement, T.P., (1997) RT3D: A Modular Computer Code for Simulation of Reactive 
Multispecies Transport in 3-Dimensional Groundwater Systems, U.S. Department of Energy. 

Drescher, E.A., R. Gavaskar, B.M. Sass, L.J. Cumming, M.J. Drescher, and T. Williamson, 
(1999) Batch and Column Testing to Evaluate Chemical Oxidation of DNAPL Source Areas, In: 
Proceeding of International Conference on Remediation of Chlorinated and Recalcitrant 
Compounds, Columbus, Ohio, pp. 425-432. 

Eilbeck, W.J., and G. Mattock, (1987) Chemical Processes in Wastewater Treatment, John Wiley 
and Sons, Toronto, Ontario. 

Gates, D.D., R.L. Siegrist, and S.R. Cline, (1995) Chemical Oxidation of Volatile and semi- 
Volatile Organic Compounds in Soil, U.S., pg. 17. 

Gates, D.D., Siegrist, R.L., and Cline, S.R., (2001) Comparison of Potassium Permanganate and 
Hydrogen Peroxide as Chemical Oxidants for Organically Contaminated Soils, Journal of 
Environmental Engineering 127(4), pp. 337-347. 

Gelhar, L.W. and M.A. Collins, (1971) General Analysis of Longitudinal Dispersion in 
Nonuniform Flow, Water Resources Research. 7(6): 1511-1521. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


75 


August 2010 



Harbaugh, A.W., E.R. Banta, M.C. Hill, and M.G. McDonald, (2000) MODFLOW-2000, the 
U.S. Geological Survey Modular Ground-Water Model - User Guide to Modularization Concepts 
and the Ground-Water Flow Process, Open-File Report 00-92, U.S. Geological Survey, pg. 121. 

Haselow, J.S., R.F. Siegrist, M. Crimi, and T. Jarosch, (2003) Estimating the Total Oxidant 
Demand for In Situ Chemical Oxidation Design, Remediation Journal 13(4), pp. 5-16. 

Hood, E.D., (2000a) Permanganate Flushing of DNAPL Source Zones: Experimental and 
Numerical Investigation, Civil and Environmental Engineering, Waterloo, University of 
Waterloo, PhD Dissertation. 

Hood, E.D., and Thomson N.R., (2000b) Numerical Simulation of In Situ Chemical Oxidation, 
In: Proceedings of International Conference on Remediation of Chlorinated and Recalcitrant 
Compounds, Columbus, Ohio, pp. 82-90. 

Huling, S.G. and B. Pivetz. 2006. “In-Situ Chemical Oxidation - Engineering Issue”. U.S. 
Environmental Protection Agency, National Risk Management Research Laboratory, R.S. Ken- 
Environmental Research Center, Ada, OK. EPA/600/R-06/072. 
(http://www.epa.gov/ada/download/issue/600R06072.pdf). 

Hutson, S.S., (2004) Estimated Use of Water in the United States in 2000, U.S. Geological 
Survey, Reston, Virginia. 

ITRC. 2005. Technical and Regulatory Guidance for In Situ Chemical Oxidation of 
Contaminated Soil and Groundwater 2nd Edition, ISCO-2. Washington, D.C.: Interstate 
Technology & Regulatory Council. http://www.itrcweb.org/gd_ISCO.asp 

Jones, L.J., (2007) The Impact of NOD Reaction Kinetic on Treatment Efficiency, Civil and 
Environmental Engineering, Waterloo, University of Waterloo, MS Thesis. 

Leung, S.W., R.J. Watts, and G.C. Miller, (1992) Degradation of Perchloroethylene by Fenton’s 
Reagent: Speciation and Pathway, Journal of Enviro nm ental Quality 21, pp. 377-381. 

Lowe, K.S., Gardner, F.G., Siegrist R.L., and Houk T.C., (1999) Field Pilot Test of In Situ 
Chemical Oxidation Through Recirculation Using Vertical Wells at the Portsmouth Gaseous 
Diffusion Plant, EPA 625-R-99-012, pp. 42-49. 

Marvin, B.K., J. Chambers, A. Leavitt, and C.G. Schreier, (2002) Chemical and Engineering 
Challenges to In Situ Permanganate Remediation, Battelle Press, Monterey, California, pp. 1127- 
1134. 

McDonald, M.G., and A.W. Harbaugh, (1988) A Modular 3-Dimensional Finite-Difference 
Ground-Water Flow Model, Techniques of Water-Resources Investigations, Book 6, US 
Geological Survey, pg. 586. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


76 


August 2010 



Mumford, K.G., N.R. Thomson, and R.M. Alien-King, (2002) Investigating the Kinetic Nature 
of Natural Oxidant Demand During ISCO, Battelle Press, Monterey, California, pp. 1383-1388. 

Mumford, K.G., C.S. Lamarche, and N.R. Thomson, (2004) Natural Oxidant Demand of Aquifer 
Materials Using the Push-Pull Technique, Journal of Enviro nm ental Engineering 130(10), pp. 
1139-1146. 

Mumford, K.G., N.R. Thomson, and R.M. Alien-King, (2005) Bench-Scale Investigation of 
Permanganate Natural Oxidant Demand Kinetics, Environmental Science and Technology 39(8), 
pp. 2835-2840. 

Oberle, D.W., and D.L Schroder, (2000) Design Considerations for In-Situ Chemical Oxidation, 
Chemical Oxidation and Reactive Barriers; Remediation of Chlorinated and Recalcitrant 
Compounds, pp. 91-99. 

Reitsma, S., and Q.L. Dai, (2001) Reaction-Enhanced Mass Transfer and Transport from Non- 
Aqueous Phase Liquid Source Zones, Journal of Contaminant Hydrology 49(1-2), pp. 49-66. 

Schnarr, M., C. Truax, G. Farquhar, E. Hood, T. Gonullu, and B. Stickney, (1998) Laboratory 
and Controlled Field Experiments Using Potassium Permanganate to Remediate 
Trichloroethylene and Perchloroethylene DNAPLs in Porous Media, Journal of Contaminant 
Hydrology 29(3), pp. 205-224. 

Siegrist, R.L., K.S Lowe, L.D. Murdoch, W.W. Slack, and T.C. Houk, (1998a) X-231A 
Demonstration of In Situ Remediation of DNAPL Compounds in Low Permeability Media by 
Soil Fracturing with Thermally Enhanced Mass Recovery or Reactive Barrier Destruction, Oak 
Ridge National Laboratory Report, ORNL/TM-13534. 

Siegrist, R.L., K.S. Lowe, L.C. Murdoch, T.L. Case, D.A. Pickering, and T.C. Houk, (1998b) 
Horizontal Treatment Barriers of Fracture-Emplaced Iron and Permanganate Particles, EPA 542- 
R-98-003, pp. 77-83. 


Siegrist, R.L., M.A. Urynowicz, and O.R. West, (2000) An Overview of In Situ Chemical 
Oxidation Technology Features and Applications, EPA 625-R-99-012, pp. 61-69. 

Siegrist, R.L., M.A. Urynowicz, O.R. West, M.L. Crimi, and K.S. Lowe, (2001) Principles and 
Practices of In Situ Chemical Oxidation Using Permanganate, Battelle Press, Columbus, OH. 

Siegrist, R.L., M.L. Crimi, B. Petri, T. Simpkin, T. Palaia, F.J. Krembs, J. Munakata-Marr, T. 
Illangasekare, G. Ng, M. Singletary, N. Ruiz. 2009. In Situ Chemical Oxidation for Groundwater 
Remediation: Site Specific Engineering & Technology Application. Final project report to the 
U.S. Environmental Security Technology Certification Program for ESTCP project ER-0623. 

Steel, E.W., and T.J. McGhee, T.J., (1979) Water Supply and Sewerage, Fifth Edition, McGraw- 
Hill, New York, New York. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


77 


August 2010 



Tompson, A.F.B., R. Aboudu, and L.W. Gelhar, (1989) Implementation of the Three 
Dimensional Turning Bands Random Field Generator, Water Resources Research 25(10), pp. 
2227-2243. 

US EPA, (1998) Field Applications of In Situ Remediation Technologies: Chemical Oxidation, 
US Environmental Protection Agency, Solid Waste and Emergency Response, EPA 542-R-98- 
008 . 

US EPA, (2007) Treatment Technologies for Site Cleanup: Annual Status Report (Twelfth 
edition), US Environmental Protection Agency, Solid Waste and Emergency Response, EPA 
542-R-07-012. 

Urynowicz, M.A., B. Balu, and U. Udayasankar, (2008) Kinetics of Natural Oxidant Demand by 
Permanganate in Aquifer Solids. Journal of Contaminant Hydrology 96(1-4), pp. 187-194. 

Van Genuchten, M. T., and P. J. Wierenga, (1974) Simulation of one dimensional solute transfer 
in porous media, Bull. 628, New Mexico State University Agricultural Experiment Station, Las 
Cruces. 

Vella, P.A., G. Deshinsky, J.E. Boll, J. Munder, and W.M. Joyce, (1990) Treatment of Low 
Level Phenols with Potassium Permanganate, Research Journal of the Water Pollution Control 
Federation 62(7), pp. 907-914. 

Vella, P.A., and B. Veronda, (1992) Oxidation of Trichloroethylene: A Comparison of 
Potassium Permanganate and Fenton’s Reagent, In: Proceedings of the Third International 
Symposium on Chemical Oxidation Technology for the Nineties, Vanderbilt University, 
Nashville, Tennessee. 

West, O.R., S.R. Cline, W.L. Holden, F.G. Gardner, and B.M. Schlosser, (1998) Field-scale Test 
of In Situ Chemical Oxidation through Recirculation, In: Proceedings of Spectrum ’98 
International conference on Nuclear and Hazardous Waste Management, Denver, Colorado, pp. 
1051-1057. 

Water Science and Technology Board (WSTC), (2004) Contaminants in the Subsurface: Source 
Zone Assessment and Remediation, Division on Earth and Life Studies, National Research 
Council of the National Academies, The National Academies Press, Washington D.C.. 

Xu, X., (2006) Interaction of Chemical Oxidants with Aquifer Materials, Civil and 
Enviro nm ental Engineering, Waterloo, University of Waterloo, PhD Dissertation. 

Yan, Y.E., and F.W. Schwartz, (1996) Oxidation of Chlorinated Solvents by Permanganate, 
Physical, Chemical, and Thermal Technologies: Oxidation Technologies, G.B. Wickramanayake 
and R. E. Hinchee eds, Battelle Press, pp. 403-408. 

Yan, Y.E., and F.W. Schwartz, (1999) Oxidative Degradation and Kinetics of Chlorinated 
Ethylenes by Potassium Permanganate, Journal of Contaminant Hydrology 37(3), pp. 343-365. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


78 


August 2010 



Zhang, H., and F.W. Schwartz, F.W., (2000) Simulating the In Situ Oxidative Treatment of 
Chlorinated Ethylenes by Potassium Permanganate, Water Resources Research 36(10), pp. 3031- 
3042. 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


19 


August 2010 



8.0 POINTS OF CONTACT 


POINT OF 
CONTACT 
Name 

ORGANIZATION 

Name 

Address 

Phone/Fax/email 

Role in Project 

Robert C. 

Borden, P.E., 
Ph.D. 

North Carolina State University 
Campus Box 7908 

Raleigh, NC 27695 

919-515-1625 
919-515-7908 (fax) 
rcborden@eos.ncsu.edu 

Principal 

Investigator 

Thomas 

Simpkin, P.E., 
Ph.D. 

CH2M HILL, Inc. 

9193 South Jamaica St 
Englwood, CO 80112-5946 

720-286-5394 
720-286-9884 (fax) 
tsimpkin@ch2m. com 

Co-Investigator 

M. Tony 

Lieberman, 

R.S.M. 

Solutions-IES, Inc. 

3722 Benson Drive 

Raleigh, NC 27609 

919-873-1060 
919-873-1074 (fax) 
tlieberman@solutions- 
ies.com 

Co-Investigator 

Erica Becvar 

HQ AFCEE/TDE Technology 
Transfer 

3300 Sidney Brooks 

Brooks City-Base, TX 78235- 
5112 

210-536-4314 
210-536-5989 (fax) 
Erica.Becvar@brooks.af. 
mil 

Contracting 

Officer’s 

Representative 

(COR) 


ESTCP Technical Report 

Design Tool for Planning Permanganate 

Injection Systems 


80 


August 2010